INTEGRATED GENETIC AND

PHENOTYPIC ASSESSMENT OF

DILATED CARDIOMYOPATHY

This dissertation is submitted to Imperial College London for the degree of

Doctor of Philosophy

by

Upasana Tayal

MA(Oxon) BMBCh MRCP(UK)

Royal Brompton Hospital Cardiovascular Research Centre

Imperial College Faculty of Medicine

August 2017

ABSTRACT

Background: (DCM) affects up to 1 in 250 individuals and is the leading global indication for heart transplantation, though a subset of patients can recover myocardial function.

Aim: To integrate clinical, genetic, and advanced imaging data to generate new insights into DCM pathobiology.

Methods and Results 1) Evaluation of the genetic architecture of DCM in 647 unrelated Caucasian patients revealed that variants in only 5 of 57 putative DCM were significantly enriched compared to >30,000 reference samples. Truncating variants in the (TTNtv) accounted for the largest genetic contribution to DCM (excess burden 14.1%, p=6.4x10-82).

2) Cardiovascular magnetic resonance (CMR) phenotype study of 716 DCM patients demonstrated that TTNtv DCM was associated with a blunted hypertrophic response (mean 2 indexed left ventricular mass, g/m ; TTNtv-/+ 91.3 vs 83.5, padjusted=0.007). Moderate alcohol excess was an environmental modifier of the TTNtv phenotype (10.0% reduction in left ventricular ejection fraction, 95% CI -16.3 to -3.8%, p=0.002).

3) In 29 patients with recent onset DCM, myocardial contractile reserve during low-dose dobutamine stress CMR was an independent predictor of LV remodelling (p=0.007). Contractile reserve assessed by a novel cine-DENSE strain CMR sequence and relative RV contractile reserve were also predictive of LV remodeling.

4) Amongst 604 DCM patients, followed up for a median of 3.9 years for the primary composite endpoint of cardiovascular mortality, major arrhythmic events and major heart failure events, the presence of TTNtv did not influence the primary outcome (TTNtv unadjusted hazard ratio 0.81 [95% CI 0.41-1.63], p=0.56). Further analyses showed no evidence that gender or CMR mid-wall fibrosis status modified the effect of TTNtv.

Conclusion These data improve our understanding of the genetic basis of DCM, provide mechanistic insight and inform risk stratification in DCM.

2 Template by Friedman & Morgan 2014 Word

This thesis is dedicated to my parents and Mayank.

Your unconditional love and support has made me everything I am today.

3 Template by Friedman & Morgan 2014 Word

THESIS OVERVIEW DIAGRAM

Research aim: Evaluate whether integrated assessment of genotype and phenotype data in DCM can improve understanding of DCM pathogenesis, inform patient stratification, and identify predictors of remodeling and outcome.

INTRODUCTION Chapter 1: General Introduction Overview of background to research question

Chapter 2: Evaluating the genetic architecture of DCM

Burden of rare variants in DCM cohort compared to population genetic variation. Purpose of this chapter is to guide subsequent phenotype analysis.

RESULTS Chapter 3: Clinical manifestations and phenotypic drivers of titin cardiomyopathy

Evaluation of the phenotype and environmental modifiers of the commonest genetic contributor to Each chapter DCM using cardiovascular magnetic resonance contains: (CMR).

• Introduction Chapter 4: Imaging predictors of cardiac • Methods remodeling • Results • Discussion Imaging sub-study to evaluate imaging predictors of • Outline of remodeling in patients with recent onset DCM, with a further work focus on contractile reserve, assessed through low- dose dobutamine CMR.

Chapter 5: The role of clinical, imaging and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Final chapter brings together chapters 2 and 3, Chapter 6: Summary discussion evaluating the prognosis of titin cardiomyopathy. What this PhD adds to the field

CONCLUSION Chapter 6: Summary overview What this PhD adds to the field and future work

4 Template by Friedman & Morgan 2014 Word

DECLARATIONS

Declaration of originality

This thesis is the result of my own work other than where duly acknowledged or

appropriately referenced. It has not been previously submitted, in part or whole, to any

university of institution for any degree, diploma, or other qualification.

Copyright declaration

The copyright of this thesis rests with the author and is made available under a Creative

Commons Attribution Non-Commercial No Derivatives licence. Researchers are free to copy,

distribute or transmit the thesis on the condition that they attribute it, that they do not use it

for commercial purposes and that they do not alter, transform or build upon it. For any reuse

or redistribution, researchers must make clear to others the licence terms of this work.

Upasana Tayal , BMBCh, MA (Oxon), MRCP (UK)

Imperial College

5 Template by Friedman & Morgan 2014 Word

ACKNOWLEDGEMENTS

I am grateful to my supervisors for giving me the opportunity to undertake this PhD. I have learnt about science, innovation and resilience in equal measure. Sanjay, thank you for believing in me from the start and for your incredible support along the way. You gave me the academic freedom to explore and helped guide me through the more challenging moments. Stuart, thank you for taking a chance on me when I barely knew the difference between genes and jeans. Your academic guidance at every stage has been invaluable and I hope it is a resource I may draw upon for years to come. James, thank you for being so generous with your time, coming on board when I needed it the most. Thanks for your mentorship and scientific rigour. I am a better scientist and I now know how to use R (sort of).

I am grateful to the MRC for funding this PhD, providing much needed peer approval of the work and setting the foundation for my academic career ahead.

I am thankful to all the staff in the BRU, Genetics and Genomics Lab, and CMR department at the Royal Brompton Hospital. Particular thanks go to Gillian Rea and Rachel Buchan, my genetics guardian angels at the start of this PhD. The genetics group is the exemplar of team science and I have learnt an enormous amount from all who have been so generous with their knowledge, including (though by no means limited to) Nicky Whiffin, Liz Edwards, Roddy Walsh, Francesco Mazzarotto, and Paul Barton. Within the BRU, thanks to Prof Pennell for ongoing academic support, both personally and for the studies in this thesis. An enormous thank you to the wonderfully dedicated army of BRU research nurses, in particular Carmen Chan, Zohreh Farzad, Sally-Ann McRae, and Annashyl West for your help with my projects. Thanks also to Steve Collins and Yasin Karafil for keeping the BRU database alive and to Geraldine Sloane for your continuous project support. Thanks also to the research fellows (especially Brian, Amrit and Zohya) for your camaraderie and for making sure that coming to work was never a chore. Within CMR, an enormous debt of gratitude to Ric Wage for your support of the remodelling study and all round CMR wisdom. Thank you also to George Mathew and Karen Symmonds for teaching me how to scan. I am lucky to have had ready access to physics expertise and have found patient teachers in Peter Gatehouse, Jenny Keegan, Andy Scott, Iain Pierce, Pedro Ferreira, and Merlin Fair, thank you all. Last but not least, I thank Simon Newsome for his never-ending statistical mentorship and support. I hope the breadth and depth of what I have learnt from you is reflected in all my academic endeavours.

A brief but heartfelt thanks must also go to my family and friends for keeping me sane over the last few years. And for the cocktails, holidays, and cuddles when said sanity was absent.

Most importantly of all, I am of course indebted to the patients and volunteers who participated in this study. Clinical research is only possible thanks to the altruism and enthusiasm of these individuals.

6 Template by Friedman & Morgan 2014 Word

CONTENTS

Abstract ...... 2

Thesis Overview Diagram ...... 4

Declarations ...... 5

Acknowledgements ...... 6

Contents ...... 7

List of Tables ...... 13

List of Figures ...... 15

List of Abbreviations and Acronyms ...... 17

Selected awards and publications ...... 19

1 Introduction: Dilated Cardiomyopathy ...... 22

1.1 What is Dilated Cardiomyopathy? ...... 22

1.1.1 Definition of Dilated Cardiomyopathy ...... 22

1.1.2 Epidemiology of Dilated Cardiomyopathy ...... 22

1.1.3 Aetiology of Dilated Cardiomyopathy ...... 23

1.2 Unmet Needs in the Assessment of Patients with Dilated Cardiomyopathy ...... 24

1.2.1 Accurate Diagnosis of Dilated Cardiomyopathy ...... 24

1.2.2 Accurate risk stratification and development of novel therapeutic targets ...... 26

1.2.3 Harnessing genetic data to improve our understanding of DCM ...... 27

1.3 Current Understanding of the Genetics of Dilated Cardiomyopathy ...... 28

1.3.1 The genetic basis of DCM ...... 28

1.3.2 Disease-gene associations in DCM ...... 38

1.4 Truncating variants in titin in DCM ...... 47

7 Template by Friedman & Morgan 2014 Word

1.4.1 Function of the titin ...... 47

1.4.2 Prevalence of titin truncating mutations in DCM ...... 48

1.4.3 Established genotype phenotype correlations in TTNtv DCM ...... 48

1.4.4 Challenges in the interpretation of TTNtv ...... 54

1.4.5 Evidence for genetic and environmental modifiers of TTNtv cardiomyopathy ...... 54

1.5 Role of Cardiovascular Magnetic Resonance in Dilated Cardiomyopathy ...... 58

1.5.1 Overview of cardiovascular magnetic resonance ...... 58

1.5.2 The role of CMR in DCM ...... 59

1.5.3 Limitations of cardiac MRI in dilated cardiomyopathy ...... 70

1.6 Remodelling and Recovery in DCM ...... 71

1.6.1 Potential for LV remodelling in DCM ...... 71

1.6.2 Situations in which LV remodelling can occur in DCM ...... 72

1.6.3 Prognostic implications of LV remodelling ...... 72

1.6.4 Predictors of LV remodelling in DCM ...... 74

1.6.5 Predictors of sustained LV recovery in DCM ...... 78

1.7 Prognosis in DCM and Risk Stratification ...... 80

1.7.1 Prognosis in DCM ...... 80

1.7.2 Established and novel risk stratification biomarkers in adult onset DCM ...... 82

1.7.3 Current risk stratification in DCM and limitations ...... 104

1.7.4 Personalised medicine in DCM ...... 110

Aims, objectives and hypotheses ...... 111

2 Evaluating the genetic architecture of dilated cardiomyopathy ...... 113

2.1 Aims and hypotheses ...... 113

2.2 Background ...... 113

2.3 Methods ...... 115

2.3.1 Study population ...... 115

2.3.2 ExAC population reference dataset ...... 117

8 Template by Friedman & Morgan 2014 Word

2.3.3 Coverage analysis between cases or controls and ExAC ...... 117

2.3.4 Gene selection ...... 117

2.3.5 Targeted sequencing ...... 119

2.3.6 General and platform specific bioinformatic analysis ...... 121

2.3.7 Variant filtering ...... 122

2.3.8 Burden testing ...... 123

2.3.9 Sample size calculation ...... 124

2.4 Results ...... 125

2.4.1 Cohort overview ...... 125

2.4.2 Coverage comparison between DCM cases and ExAC ...... 127

2.4.3 Excluded variants ...... 130

2.4.4 Burden of rare genetic variation in DCM patients compared to the ExAC cohort ...... 131

2.4.5 Burden of rare genetic variation in healthy volunteers compared to the ExAC cohort

...... 139

2.4.6 Proportion of DCM cases with identifiable genetic variant ...... 142

2.5 Discussion ...... 143

2.5.1 Limitations of this study ...... 150

2.6 Summary ...... 153

2.7 Outline of further work ...... 154

2.8 Acknowledgements ...... 154

3 The clinical manifestations and phenotypic drivers of titin cardiomyopathy ...... 155

3.1 Aims and Hypotheses ...... 155

3.2 Background ...... 155

3.3 Methods ...... 158

3.3.1 Study Population ...... 158

3.3.2 Ethics ...... 158

3.3.3 Titin truncating variant curation for phenotype analysis ...... 158

9 Template by Friedman & Morgan 2014 Word

3.3.4 DCM phenotyping ...... 163

3.3.5 Clinical Data ...... 165

3.3.6 Statistical Analysis ...... 169

3.4 Results ...... 171

3.4.1 Cohort Recruitment ...... 171

3.4.2 Patient demographics ...... 172

3.4.3 Overall DCM phenotype ...... 174

3.4.4 Curation of titin truncating variants ...... 180

3.4.5 TTNtv cardiomyopathy phenotype ...... 193

3.4.6 Alcohol as a phenotypic modifier of TTNtv cardiomyopathy ...... 207

3.4.7 Other modifiers of TTNtv cardiomyopathy ...... 222

3.5 Discussion ...... 225

3.5.1 Strengths of this study ...... 230

3.5.2 Limitations of this study ...... 231

3.6 Summary ...... 232

3.7 Outline of further work ...... 233

3.8 Acknowledgements ...... 233

4 Imaging predictors of cardiac remodelling ...... 234

4.1 Aims and hypotheses ...... 234

4.2 Background ...... 235

4.3 Methods ...... 240

4.3.1 Study overview ...... 240

4.3.2 Study cohort, inclusion and exclusion criteria ...... 240

4.3.3 Baseline CMR protocol ...... 242

4.3.4 Follow up imaging ...... 254

4.3.5 Sample size calculation and statistical analysis ...... 254

4.4 Results ...... 257

10 Template by Friedman & Morgan 2014 Word

4.4.1 Cohort size and loss to follow up ...... 257

4.4.2 Baseline demographics and CMR findings ...... 258

4.4.3 Safety of dobutamine ...... 262

4.4.4 Heart rate and blood pressure response to dobutamine stress ...... 262

4.4.5 Baseline LV contractile reserve ...... 264

4.4.6 Baseline LV strain ...... 267

4.4.7 Baseline interstitial fibrosis ...... 273

4.4.8 Evaluating absolute contractile reserve as a predictor of LV remodelling ...... 277

4.4.9 Other predictors of LV remodelling ...... 284

4.4.10 Mechanistic basis of contractile reserve ...... 288

4.5 Discussion ...... 289

4.5.2 Strengths of this study ...... 296

4.5.3 Limitations of this study ...... 298

4.6 Summary ...... 302

4.7 Outline of further work ...... 303

4.8 Acknowledgements ...... 303

5 The role of clinical, imaging, and genetic data in predicting clinical outcomes in

dilated cardiomyopathy ...... 304

5.1 Aims and hypotheses ...... 305

5.2 Background ...... 305

5.3 Methods ...... 306

5.3.1 Study cohort ...... 306

5.3.2 Variable definition and analysis ...... 306

5.3.3 Endpoints and adjudication of events ...... 307

5.3.4 Statistical analysis ...... 310

5.4 Results ...... 313

5.4.1 Overview ...... 313

11 Template by Friedman & Morgan 2014 Word

5.4.2 Building the baseline model to predict the primary end point ...... 316

5.4.3 Analysis of the association between TTNtv and the primary end point ...... 325

5.4.4 Sensitivity analyses for the association between TTNtv and the primary endpoint . 331

5.4.5 Non-titin genetic variants and clinical outcomes: primary endpoint ...... 336

5.4.6 Analysis of the association between TTNtv and secondary outcome end points ...... 338

5.5 Discussion ...... 340

5.5.2 Strengths of this study ...... 345

5.5.3 Limitations of this study ...... 345

5.6 Summary ...... 349

5.7 Outline of further work ...... 349

5.8 Acknowledgements ...... 349

6 Summary: What this thesis adds to the field and future work ...... 351

7 References ...... 356

8 Appendices ...... 373

8.1 Gene panels ...... 373

8.2 Standard Operating Protocol for Date of Clinical Diagnosis of DCM ...... 387

8.3 CMR safety checklist ...... 388

8.4 Endpoint definitions ...... 389

8.5 Permissions ...... 392

12 Template by Friedman & Morgan 2014 Word

LIST OF TABLES

Table 1-1: Examples of identifiable causes of dilated cardiomyopathy ...... 23 Table 1-2: Genes implicated in monogenic dilated cardiomyopathy ...... 30 Table 1-3: Composition of the ExAC dataset ...... 36 Table 1-4: Genes linked to DCM, annotated by year of report in HGMD ...... 39 Table 1-5: Summary of key genotype-phenotype studies of TTNtv DCM ...... 53 Table 1-6: Biomarkers linked to heart failure classified by biological function ...... 98 Table 1-7: Current guidelines for primary prevention ICD implantation ...... 104 Table 2-1: Genes with variants reported to be associated with DCM in HGMD ...... 118 Table 2-2: Genes and transcript linked to DCM included on Tru Sight Cardio ...... 119 Table 2-3: Coverage comparison between DCM and ExAC datasets ...... 128 Table 2-4: Poor quality variants excluded from burden analysis ...... 130 Table 2-5: Burden of rare truncating variants in DCM genes in DCM cohort ...... 134 Table 2-6: Burden of rare non-truncating variants in DCM genes in DCM cohort ...... 137 Table 2-7: Burden of rare variants in RBM20 in DCM cohort compared to volunteers ...... 139 Table 2-8: Burden of truncating and non-truncating variants in TTN in DCM patients ...... 140 Table 3-1: Genes and variant classes selected for variant curation ...... 160 Table 3-2: Titin meta-transcript details ...... 163 Table 3-3: Data collected on DCM Events form as baseline phenotyping of cohort ...... 166 Table 3-4: Baseline demographics in DCM cohort ...... 173 Table 3-5: Baseline CMR phenotype in DCM cohort ...... 178 Table 3-6: Sequencing platform and assay used in DCM cohort ...... 182 Table 3-7: List of truncating variants in TTN gene ...... 186 Table 3-8: Baseline demographics in DCM patients stratified by TTNtv status ...... 194 Table 3-9: Baseline CMR variables in DCM cohort stratified by TTNtv status ...... 196 Table 3-10: Variables evaluated on univariable linear regression as predictors of left ventricular mass ...... 198 Table 3-11: Multivariable regression analysis evaluating predictors of left ventricular mass ...... 200 Table 3-12: Sarcomeric variant sensitivity analysis ...... 203 Table 3-13: Baseline demographics and cardiac phenotype, stratified by alcohol status ..... 209 Table 3-14: Results of univariable and multivariable linear regression evaluating predictors of LVEF ...... 212 Table 3-15: Results of nested ANOVA comparing baseline models to predict LVEF...... 214 Table 3-16: Results of univariable and multivariable linear regression analysis evaluating the effect of TTNtv and alcohol excess on LVEF ...... 216 Table 4-1 CMR baseline study protocol ...... 243 Table 4-2: Baseline demographics and CMR findings in cohort stratified by diagnosis ...... 261 Table 4-3: Linear regression models evaluating the effect of baseline variables on LV absolute contractile reserve...... 266 Table 4-4: Peak global long axis (HLA, VLA) and SAX strain in DCM patients and volunteers ...... 270 Table 4-5: Absolute difference in peak global long axis (HLA, VLA) and SAX strain in DCM patients and healthy volunteers ...... 273 Table 4-6: Extra-cellular volume fraction (ECV) in DCM patients and healthy volunteers . 274 Table 4-7: Univariable ANCOVA regression evaluating predictors of follow up LVEF. .... 279 Table 4-8: Multivariable ANCOVA regression analysis evaluating absolute contractile

13 Template by Friedman & Morgan 2014 Word

reserve ...... 281 Table 4-9: Contractile reserve sensitivity analysis ...... 283 Table 4-10: Contractile reserve assessed by myocardial strain as predictor of follow up LVEF...... 284 Table 4-11: Results of ANCOVA regression analysis evaluating baseline indices of strain as predictors of follow up LVEF...... 287 Table 5-1: Variables evaluated in survival analysis ...... 306 Table 5-2: Table showing the number of patients meeting each endpoint ...... 314 Table 5-3: Composition of events occurring during follow up leading to primary composite endpoint ...... 315 Table 5-4: Baseline demographics in the outcome cohort ...... 317 Table 5-5: Results of univariable Cox proportional hazard modeling ...... 319 Table 5-6: Adjusted baseline Cox proportional hazard model predicting the primary endpoint ...... 321 Table 5-7: Comparison of Cox models ...... 322 Table 5-8: Results of testing proportionality assumption of baseline model ...... 323 Table 5-9: Adjustment of Cox model for the time dependent covariate of cardiac device implantation ...... 333 Table 5-10: Sensitivity analysis: VT ...... 334 Table 5-11: Association between TTNtv and the primary endpoint, adjusted for the baseline Cox model in the subset of patients with complete serum creatinine data...... 335 Table 5-12: Numbers of patients with rare variants (MAF <0.0001) in non-titin DCM genes meeting the primary endpoint ...... 336 Table 5-13: Association between non-TTNtv genetic variants and the primary endpoint .... 337 Table 8-1: Definition of cardiovascular death ...... 389

14 Template by Friedman & Morgan 2014 Word

LIST OF FIGURES

Figure 1-1: Model of types of genetic susceptibility to disease ...... 29 Figure 1-2: Overview of NGS workflow ...... 34 Figure 1-3: Putative model of genetic and environmental modifiers of TTNtv ...... 55 Figure 2-1: Overview of methods for variant curation for burden testing ...... 123 Figure 2-2: Simulated power calculation for burden testing ...... 125 Figure 2-3: Overview of cohorts in burden testing ...... 126 Figure 2-4: Coverage plot ...... 127 Figure 2-5: Excess burden of protein altering variants ...... 132 Figure 2-6: Pie chart of excess burden (%) in significantly enriched genes (p<0.0009) in DCM cohort ...... 132 Figure 2-7: Burden of rare truncating variants in DCM compared to ExAC ...... 133 Figure 2-8: Burden of rare non-truncating variants in DCM compared to ExAC ...... 136 Figure 2-9: Burden of rare protein altering variants in volunteers compared to ExAC ...... 141 Figure 2-10: Bar chart showing percentage of DCM cases with identified genetic variant .. 142 Figure 3-1: Outline of DCM cohort recruitment for phenotype study ...... 171 Figure 3-2: Interval between date of diagnosis and baseline date ...... 174 Figure 3-3: Histograms showing the distribution of left and right ventricular function ...... 176 Figure 3-4: CMR phenotype of DCM cohort compared to controls ...... 177 Figure 3-5: CMR phenotype correlation plot ...... 179 Figure 3-6: Flowchart outlining filtering steps to curate TTNtv for phenotype analysis ...... 181 Figure 3-7: Percentage coverage of TTN at 30x ...... 183 Figure 3-8: IGV output of individual with 2 protein altering variants in TTN ...... 184 Figure 3-9: Position of TTNtv variants according to protein domains and PSI ...... 192 Figure 3-10: Cardiac phenotype in probands and affected relatives ...... 193 Figure 3-11: TTNtv hypertrophy phenotype ...... 197 Figure 3-12: LV mass and LV end diastolic volume ...... 200 Figure 3-13: Forest plot showing the predictors of left ventricular mass ...... 201 Figure 3-14: Model checking of final linear regression model predicting LV mass ...... 202 Figure 3-15: Plots show the relationship between TTNtv location and cardiac endophenotypes assessed by CMR ...... 205 Figure 3-16: Cronos: Plots show the relationship between TTNtv location and cardiac endophenotypes assessed by CMR ...... 206 Figure 3-17: Self reported weekly alcohol consumption (units) across study cohort ...... 207 Figure 3-18: LVEF in patients with TTNtv and alcohol excess ...... 210 Figure 3-19: Forest plot showing results of final linear regression model predicting LVEF and the effects of TTNtv and alcohol ...... 217 Figure 3-20: Model checking of final linear regression model predicting LVEF with the TTNtv and alcohol interaction ...... 218 Figure 3-21: Log2 transformation of alcohol consumption plotted against LVEF ...... 220 Figure 3-22: Weekly alcohol consumption grouped into categorical variables and plotted against LVEF ...... 221 Figure 3-23: Forest plot of estimate and 95% confidence intervals of TTNtv:Alcohol interaction term predicting LVEF ...... 221 Figure 3-24: LVEF stratified by TTNtv and hypertension status ...... 222 Figure 3-25: Forest plot showing results of final linear regression model predicting LVEF and the effects of TTNtv and hypertension ...... 223

15 Template by Friedman & Morgan 2014 Word

Figure 3-26: LVEF and age at study recruitment in patients with TTNtv ...... 224 Figure 4-1: DCM Remodelling study overview ...... 240 Figure 4-2: Example of threshold based measurement of cardiac volumes ...... 244 Figure 4-3: Schematic of dobutamine infusion stages ...... 245 Figure 4-4: Examples of defining single line contour regions of interest for the assessment of long axis strain ...... 250 Figure 4-5: Examples of defining regions of interest for the assessment of short axis strain 251 Figure 4-6: Figure demonstrating zonal excitation ...... 252 Figure 4-7: Diagram outlining the relationship between the exposure variable (contractile reserve) and the outcome variable (follow up LVEF) ...... 256 Figure 4-8: Overview of cohort recruitment and completion to follow up ...... 258 Figure 4-9: Baseline prognostic medication in DCM cohort ...... 259 Figure 4-10: Maximum change in systolic blood pressure (SBP) and heart rate during dobutamine infusion ...... 263 Figure 4-11: Maximal change in systolic blood pressure (SBP) and heart rate during dobutamine infusion ...... 263 Figure 4-12: Comparison of the absolute change in LVEF from baseline to peak dobutamine stress ...... 264 Figure 4-13: Absolute contractile reserve ...... 265 Figure 4-14: Plot of relative LV contractile reserve in DCM patients and controls ...... 267 Figure 4-15: Cine DENSE analysis ...... 268 Figure 4-16: Summary strain-time curves for DCM and control cohort ...... 269 Figure 4-17: Relationship between left ventricular ejection fraction and contour strains ..... 271 Figure 4-18: Correlation matrix ...... 272 Figure 4-19: Extracellular volume fraction (ECV) across cohort ...... 274 Figure 4-20: Relationship between ECV and LVEF ...... 275 Figure 4-21: LVEF change at 1 year ...... 276 Figure 4-22: Regression models evaluating absolute contractile reserve ...... 282 Figure 4-23: Absolute RV contractile reserve in DCM patients compared to controls ...... 285 Figure 4-24: Right and left ventricular contractile reserve ...... 285 Figure 4-25: Relative RV contractile reserve in DCM patients and healthy volunteers ...... 286 Figure 4-26: Contractile reserve and ECV ...... 288 Figure 5-1: Outline of variable selection for the baseline Cox proportional hazard model .. 312 Figure 5-2: Histogram of follow up time in original dataset ...... 314 Figure 5-3: Graphical diagnostic of the baseline Cox regression model ...... 324 Figure 5-4: Survival curves comparing freedom from the primary endpoint in patients with and without TTNtv ...... 325 Figure 5-5: Survival curves showing freedom from the primary endpoint comparing patients with and without TTNtv stratified by gender ...... 327 Figure 5-6: Survival curves showing freedom from primary endpoint in subset of patients with TTNtv, stratified by a history of alcohol excess ...... 328 Figure 5-7: Survival curves showing outcome in DCM cohort stratified by a baseline history of alcohol excess ...... 328 Figure 5-8: Freedom from primary endpoint in DCM stratified by TTNtv and baseline history of mid wall fibrosis LGE ...... 329 Figure 5-9: Forest plot of Cox proportional hazard modelling of TTNtv interactions ...... 331 Figure 5-10: Survival curve for patients meeting the primary endpoint stratified by LMNA variant status ...... 337 Figure 5-11: Freedom from secondary endpoints stratified by TTNtv status ...... 339 Figure 8-1: CMR safety checklist ...... 388

16 Template by Friedman & Morgan 2014 Word

LIST OF ABBREVIATIONS AND ACRONYMS

1000GP; 1000 Genomes Project ACEi; Angiotensin converting enzyme inhibitor ACMG; American College of Medical Genetics ARVD/ARVC: arrhythmogenic right ventricular dysplasia/cardiomyopathy BAG3; BCL2 Associated Athanogene 3 gene BNP; B-type natriuretic peptide BSA; body surface area CMR; cardiovascular magnetic resonance CRT-P/D; cardiac resynchronization therapy –pacemaker/defibrillator dbNSFP; database for non synonymous SNPs’ functional predictions DCM; dilated cardiomyopathy DENSE; displacement encoding with simulated echoes DSP; gene ECV; extracellular volume fraction EPS; electrophysiology studies ESC; European Society of Cardiology ESP; NHLBI GO Exome Sequencing Project ExAC; Exome Aggregation Consortium FLNC; C gene FOV; field of view FWHM; full-width at half maximum gnomAD; Genome Aggregation Database GWAS; genome wide association study HCM; hypertrophic cardiomyopathy HLA; horizontal long axis HVOL; healthy volunteer ICD; implantable cardiac defibrillator LA; left atrial LGE; late gadolinium enhancement LMNA; A/C gene LVEDd; left ventricular end-diastolic diameter LVEDVi; indexed left ventricular end diastolic volume LVEF; left ventricular ejection fraction LVESVi; indexed left ventricular end systolic volume LVMi; indexed left ventricular mass LVOT; left ventricular outflow tract LVRR: left ventricular reverse remodelling LVSVi; indexed left ventricular stroke volume MAF; minor allele frequency MMPs; Matrix metalloproteinases

17 Template by Friedman & Morgan 2014 Word

mTOR; mammalian target of rapamycin MYBPC3; binding protein C gene MYH7; Beta myosin heavy chain gene NGS; next-generation DNA sequencing NSVT; non-sustained ventricular tachycardia NTproBNP; N-terminal pro BNP PKA; cyclic-AMP–dependent protein kinase PKG; cyclic-GMP–dependent protein kinase PLN; gene PSI; percentage spliced in RBM20; RNA Binding protein 20 gene RVEDVi; indexed right ventricular end diastolic volume RVEF; right ventricular ejection fraction RVESVi; indexed right ventricular end systolic volume RVOT; right ventricular outflow tract RVSVi; indexed right ventricular stroke volume SAX; short axis SBP; systolic blood pressure SCD; sudden cardiac death SD; standard deviation SNP; single nucleotide polymorphism SSFP; Steady-state free precession MRI STIR; T2 weighted Short-Tau Inversion Recovery sequence TIMPs; tissue inhibitors of matrix metalloproteinases TNNC1; cardiac C gene TNNI3; cardiac gene TNNT2; cardiac gene TTNtv; truncating variant in titin gene TWA; microvolt T wave alternans VCF; variant call format VF; ventricular fibrillation VLA; vertical long axis VT; ventricular tachycardia VUS; variants of uncertain significance WES; whole exome sequencing WGS; whole genome sequencing ZBTB17; zinc-finger and BTB domain-containing protein 17 gene

18 Template by Friedman & Morgan 2014 Word

SELECTED AWARDS AND PUBLICATIONS

Prizes related to work in this thesis

American College of Cardiology Mar 2017: Winner of the American College of Cardiology Young Investigator Award (Clinical) (ACC, Washington, USA).

London Cardiovascular Society Feb 2017: Winner of the London Cardiovascular Society JRSM Cardiovascular Disease Young Investigator of the Year Award (The Medical Society of London).

British Society of Cardiovascular Magnetic Resonance Mar 2017: Second place in British Society of Cardiac MRI (BSCMR) Young Investigator Award and second place in BSCMR e-poster competition (Manchester).

Imperial College London July 2017: Highly commended oral presentation at the National Heart Lung Institute Post Graduate Research Day (Imperial College London). April 2017: Awarded 2nd place in the Imperial Graduate School 3 minute thesis competition (Imperial College London).

British Cardiac Society June 2017: Evaluation of titin cardiomyopathy in patients with dilated cardiomyopathy reveals a blunted hypertrophic response, an early arrhythmic risk and a significant interaction with alcohol. Poster, British Cardiac Society Awarded BHF/BCS prize for the highest ranked abstract in the Valve Disease/Pericardial Disease/Cardiomyopathy category. June 2015: Comprehensive Assessment Of Rare Genetic Variation In Dilated Cardiomyopathy Genes In Patients And Controls. Poster, British Cardiac Society Awarded BHF/BCS prize for the highest ranked abstract in the Valve Disease/Pericardial Disease/Cardiomyopathy category.

Publications related to work in this thesis

Original articles 1. Tayal U, Newsome S, Buchan R, Whiffin N, Walsh R, Barton PJ, Ware JS, Cook SA, Prasad SK: Truncating Variants in Titin Independently Predict Early Arrhythmias in Patients With Dilated Cardiomyopathy. J Am Coll Cardiol 2017, 69:2466-2468.

2. Halliday BP, Gulati A, Ali A, Guha K, Newsome S, Arzanauskaite M, Vassiliou VS, Lota A, Izgi C, Tayal U, et al: Association Between Midwall Late Gadolinium Enhancement and Sudden Cardiac Death in Patients With Dilated Cardiomyopathy and Mild and Moderate Left Ventricular Systolic Dysfunction. Circulation 2017, 135:2106-2115.

19 Template by Friedman & Morgan 2014 Word

3. Schafer S, de Marvao A, Adami E, Fiedler LR, Ng B, Khin E, Rackham OJ, van Heesch S, Pua CJ, Kui M, Walsh R, Tayal U et al: Titin-truncating variants affect heart function in disease cohorts and the general population. Nat Genet 2017, 49:46- 53.

Reviews, editorials, or invited publications

1. Tayal U, Prasad S, Cook SA: Genetics and genomics of dilated cardiomyopathy and systolic heart failure. Genome Med 2017, 9:20.

2. Tayal U, Cook SA: Truncating Variants in Filamin C: The Challenges of Genotype- Phenotype Correlations in Cardiomyopathies. J Am Coll Cardiol 2016, 68:2452-2453.

Abstracts related to work in this thesis

1. Tayal, U., Buchan, R., Whiffin, N., Newsome, S., Walsh, R., Barton, P., . . . Prasad, S. (2017). Evaluation Of Titin Cardiomyopathy In Patients With Dilated Cardiomyopathy Reveals A Blunted Hypertrophic Response, An Early Arrhythmic Risk And A Significant Interaction With Alcohol. Heart 103 (Pp. A95). 2. Tayal, U., Newsome, S., Buchan, R., Whiffin, N., Walsh, R., Ware, J., . . . Cook, S. A. (2016). Defining titin cardiomyopathy: genotype- phenotype correlations in a large prospective cohort of dilated cardiomyopathy patients. European Heart Journal Vol. 37 (pp. 365). 3. Halliday, B. P., Ali, A., Gulati, A., Newsome, S., Tayal, U., Lota, A., . . . Prasad, S. K. (2016). Gender differences in the natural history and outcome of dilated cardiomyopathy. European Heart Journal Vol. 37 (pp. 325-326). 4. Tayal, U., Newsome, S., Buchan, R., Whiffin, N., Walsh, R., Ware, J., . . . Prasad, S. K. (2016). Genetic determinants of arrhythmia in dilated cardiomyopathy. European Heart Journal Vol. 37 (pp. 206). 5. Lota, A., Wassall, R., Tsao, A., Shakur, R., Tayal, U., Halliday, B., . . . Prasad, S. K. (2016). Prevalence and prognostic significance of right ventricular systolic function assessed by CMR in patients with suspected acute myocarditis. European Heart Journal Vol. 37 (pp. 1365-1366). 6. Halliday BP, Gulati A, Ali A, Guha K, Arzanauskaite M, Newsome S, Lota A, Tayal U,. . . Prasad, S. K. (2016). Risk stratification of mild-to-moderate phenotypes of dilated cardiomyopathy - the role of mid-wall fibrosis. European Heart Journal Vol. 37 (pp. 198-199). 7. Halliday, B. P., Ali, A., Gulati, A., Newsome, S., Tayal, U., Wage, R., . . . Prasad, S. K. (2016). The natural history of non-ischaemic dilated cardiomyopathy diagnosed after the age of 65 years of age. European Heart Journal Vol. 37 (pp. 1324). 8. Tayal, U., Buchan, R., Whiffin, N., et al. (2016). Clinical and genetic characteristics of familial dilated cardiomyopathy in a large UK prospective cohort. Heart 102(Suppl 6): A103 9. Tayal, U., Buchan, R., Whiffin, N., et al (2016). Effects of truncating variants in titin on cardiac phenotype and left ventricular remodelling in dilated cardiomyopathy. Heart 102(Suppl 6): A102 10. Tayal, U., Newsome, S., Buchan, R., et al (2016). The presence of a truncating mutation in titin independently associates with arrhythmic burden in patients with

20 Template by Friedman & Morgan 2014 Word

dilated cardiomyopathy. European Journal of Heart Failure 18 (Suppl 1):156 11. Tayal, U., Newsome S, Wage R, et al (2016). Evaluation of CMR predictors of right ventricular remodelling in dilated cardiomyopathy. Journal of Cardiovascular Magnetic Resonance 2016 18(Suppl 1):P281 12. Tayal, U., Newsome S, Wage R, et al (2016). Evaluating the significance of left ventricular midwall fibrosis detected by late gadolinium enhancement imaging on left ventricular functional remodelling in dilated cardiomyopathy. Journal of Cardiovascular Magnetic Resonance 2016 18(Suppl 1):P283 13. Scott AD, Tayal U, Nielles-Vallespin S, et al (2016). Accelerating cine DENSE using a zonal excitation. Journal of Cardiovascular Magnetic Resonance 2016 18(Suppl 1):O50 14. Tayal, U., Mazzarotto, F., Buchan, R., et al. (2015). Comprehensive Assessment of Rare Genetic Variation in Dilated Cardiomyopathy Genes in Patients and Controls. Heart 101(Suppl 4): A41-A42.

Invited talks

• CMR and Cardiomyopathy at the American College of Cardiology Conference March 2017, Washington, USA. • ‘Genetics of Dilated Cardiomyopathy: What’s new?’ Invited research presentation at the Institute of Cardiovascular Medicine and Science Symposium- The Future of Cardiovascular Medicine. National Heart and Lung Institute, London. September 2016 • ‘Big hearts and giant genes- why size does matter’. Invited lecture, Eton College Medical Society. Jan 2016. • ‘Titin cardiomyopathy’. Invited lecture at the 9th Cardiomyopathy Workshop, Imperial College. Dec 2014.

Book Chapters

• Chapter on Dilated Cardiomyopathy and Myocarditis for the Diagnosis and Management of Adult Congenital Heart Disease, Gatzoulis, Webb and Daubeney, 3rd edition [Churchill Livingstone]. • Chapter on Dilated Cardiomyopathy for Cardiovascular Magnetic Resonance- Manning and Pennell, 3rd edition [Elsevier]. (in press)

21 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1 INTRODUCTION: DILATED

CARDIOMYOPATHY

1.1 What is Dilated Cardiomyopathy?

1.1.1 Definition of Dilated Cardiomyopathy

Dilated cardiomyopathy (DCM) is a disease of the myocardium characterised by left

ventricular dilatation and systolic impairment in the absence of coronary artery disease or

abnormal loading conditions, such as hypertension or valvular heart disease, sufficient to

explain the observed myocardial abnormality (1-3).

1.1.2 Epidemiology of Dilated Cardiomyopathy

The landmark population study, the Olmsted County study, performed between 1975 and

1984, estimated the prevalence of DCM to be in the region of 1 in 2,700 individuals (4). This

however was prior to the widespread availability of echocardiography. The prevalence of

hypertrophic cardiomyopathy in the same study was estimated to be ~1 in 5000, which has

since been shown to be a gross under-estimate. Therefore, in the absence of large

contemporary population studies, estimated DCM prevalence has recently been revised to 1

in 250 individuals(5). DCM is a leading cause of heart failure and is the primary global

indication for heart transplantation (1,6,7).

22 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.1.3 Aetiology of Dilated Cardiomyopathy

The major identifiable causes of DCM include genetic or syndromic forms, infectious

diseases, drugs and toxins, endocrine disorders, inflammatory conditions, and nutritional

deficiencies (examples in Table 1-1). DCM is classified as idiopathic in 50% of cases when

non-genetic identifiable causes have been excluded. The genetic basis of DCM is reviewed in

detail in Section 1.3.

Table 1-1: Examples of identifiable causes of dilated cardiomyopathy

Cause Examples

Over 60 genes reported to be associated with DCM including TTN (up to 25%), Genetic and LMNA, MYH7 and TTNT2 (<5%). syndromic Duchenne Muscular Dystrophy, Barth Syndrome

Viral: Coxsackie, HIV, Influenza, Adenovirus, Cytomegalovirus, Varicella, Hepatitis, Ebstein-Barr, Echovirus, Parvovirus Bacterial: Streptococci, Mycobacteria Infectious Spirochetal: Lyme disease, Syphilis Fungal: Histoplasmosis, Cryptococcocis, Parasitic: Toxoplasmosis, Trypanosomiasis, Schistosomiasis

Chemotherapeutic agents including anthracyclines and cyclophosphamide Drugs Antiretroviral drugs including Zidovudine, Other, e.g. Phenothiazines, Chloroquine, Clozapine

Toxins Alcohol, Cocaine, Amphetamines, Cobalt, Lead, Mercury Nutritional deficiencies and Thiamine, Selenium, Carnitine, Niacin electrolyte Hypocalcaemia, hypophosphatemia, Ureamia disturbances

Hypo- or hyper-thyroidism, Diabetes mellitus, Cushing’s syndrome, Endocrine Phaeochromocytoma, Growth hormone excess or deficiency

Inflammatory and Systemic Lupus Erythematosis, Scleroderma, Rheumatoid arthritis, Autoimmune Autoimmune myocarditis, Dermatomyositis

Other Tachycardia induced cardiomyopathy, Pregnancy

23 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.2 Unmet Needs in the Assessment of Patients with Dilated

Cardiomyopathy

There are a number of areas of unmet need relating to the assessment and management of

patients with DCM, including accurate diagnosis, risk stratification, and development of

novel therapeutics to improve outcomes.

1.2.1 Accurate Diagnosis of Dilated Cardiomyopathy

DCM is an imaging and clinical diagnosis. According to the 2008 European Society of

Cardiology (ESC) guidelines, the diagnosis of DCM requires the presence of imaging

evidence of left ventricular dilation and systolic impairment, in the absence of clinical

evidence of abnormal loading conditions which would otherwise account for the imaging

phenotype and after the exclusion of coronary artery disease sufficient to cause global

systolic impairment(8).

The most commonly used imaging modality is echocardiography, in which the criteria for left

ventricular dilation and impairment are left ventricular end-diastolic diameter (LVEDd) >

117% of that predicted for age and body surface area (BSA) and left ventricular ejection

fraction (LVEF) < 45% and/or fractional shortening (FS) < 25% respectively(9).

However, these criteria are not definitive, and patients who do not meet these

echocardiographic criteria may still have DCM, as evidenced by clinically significant

myocardial disease on cardiovascular magnetic resonance (CMR) or endomyocardial

biopsy(10). As our understanding of the genetic basis of DCM increases, it is now

recognized that mutation carriers have intermediate phenotypes that do not meet standard

24 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

disease definitions. Therefore in 2016, a revision to the ESC classification was proposed to

include a category of hypokinetic non-dilated cardiomyopathy, to recognise the presence of

left ventricular impairment (LVEF <45%) in the absence of dilation as being on the same

disease spectrum as dilated cardiomyopathy(2). In this revision, the authors also proposed a

broader definition for DCM: ‘Left ventricular or biventricular systolic dysfunction and

dilatation that are not explained by abnormal loading conditions or coronary artery

disease’(2). Crucially, the revised definition accepts abnormal LVEF measured using any

modality and defines LV dilation as LV end diastolic volumes or diameters greater than 2

standard deviations from age, gender and BSA adjusted nomograms(2).

These revisions acknowledge that historically there has been a marked heterogeneity in the

assessment of DCM. This has implications not only for the clinical assessment of patients,

but also in the interpretation of research studies with ‘DCM’ cohorts. Patients with DCM and

mild-moderate LV impairment may have been excluded from previous studies. Furthermore,

echocardiographic assessment is also limited by the need for a suitable anatomical window

and poor inter- and intra observer variability, and also does not provide tissue

characterisation (11). This could lead to misclassification of patients, particularly in

conditions with phenotypic overlap, for example arrhythmogenic ventricular

cardiomyopathy, or in the mid range of LVEF (40-50%) where the established magnitude of

measurement variability in echocardiography could affect whether a diagnosis of DCM is

made.

In these situations, CMR imaging offers great incremental value in the diagnosis of DCM,

enabling accurate assessment of cardiac volumes and function, as well as providing detailed

tissue characterisation. As yet, there are no large scale studies of patients with DCM

25 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

confirmed by CMR.

1.2.2 Accurate risk stratification and development of novel therapeutic

targets

Despite improvements in optimal medical therapy, DCM remains associated with adverse

long term outcomes, with a 20% 5-year mortality rate(12,13). A key challenge in the

management of DCM patients is improved risk stratification, both the early identification of

the ‘at risk’ patient, as well as targeting of existing therapy to individuals likely to receive the

greatest benefit.

The current treatment paradigm in DCM is for all patients to broadly receive the same

therapy (ACE (Angiotensin converting enzyme) inhibitor, beta blocker, aldosterone

antagonist), irrespective of underlying aetiology or genetic basis. Novel heart failure

therapies such as the angiotensin-neprilysin inhibitor have recently emerged but it is too soon

to be able to evaluate long term prognostic benefit in DCM[11]. There is a pressing need to

identify novel therapeutic targets in DCM, which may be achieved through improved

understanding of the mechanistic basis of disease. Until then, tailoring current therapies more

precisely to an individual patient is an area of much promise, but has yet to be explored in

detail [12].

Risk stratification tools in DCM are limited and largely based on qualitative clinical data,

imaging features and biochemical markers, much of which reflect changes observed late in

the disease course. Faced with these difficulties the ideal risk assessment tool would be one

that identifies patients at risk of heart failure prior to overt disease, at a time when a

26 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

preventative intervention could be used to avoid disease onset. Genetics offers one such

approach.

1.2.3 Harnessing genetic data to improve our understanding of DCM

There have been major advances in DNA sequencing technologies over recent years, which

have enabled the widespread application of DNA sequencing to cardiomyopathy cohorts(14).

This has led to a rapid increase in the number of genes reported to be associated with

DCM(5).

At an even more rapid pace, DNA sequencing at scale has been applied in very large

population cohorts, such as those included in the Exome Aggregation Consortium (ExAC)

data set (15) (now renamed the Genome Aggregation Database (gnomAD) to reflect the

inclusion of genome sequencing data). Against this background, understanding which genes

and variants are of importance for a patient with DCM is a challenge for the clinician.

The majority of the gene-DCM associations have no clear genotype-phenotype or genotype-

outcome correlations. How this increasing genetic knowledge informs our understanding of

the pathogenesis of DCM or changes clinical management is not yet fully established.

27 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.3 Current Understanding of the Genetics of Dilated

Cardiomyopathy

1.3.1 The genetic basis of DCM

The proportion of DCM cases with a familial basis is between 20-30%, though up to 60% has

been suggested (16). In familial DCM, up to 40% of cases may have an identifiable genetic

basis (5). Inheritance of DCM is most commonly autosomal dominant, although autosomal

recessive, X-linked and mitochondrial inheritance have also been reported, particularly in

pediatric populations (17).

As a more critical evaluation of the genes linked to DCM continues and genes or variants are

discounted, the percentage of DCM with a monogenic basis may fall (14,18). Higher

estimates of sensitivity for genetic testing have been reported (from 46-73% in one study

(19)) but these estimates are likely confounded by insufficient control for population

variation in the genes studied. After excluding non-genetic causes, the low genetic detection

rate in the presence of familial DCM may be attributed to an incomplete understanding of the

genetic basis of DCM. One possibility is that DCM may have a more complex oligogenic

genetic inheritance, where genetic susceptibility may be determined by the cumulative effect

of multiple genetic variants, as opposed to a near Mendelian inheritance driven by a strong

monogenic component (Figure 1-1)(20).

28 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Figure 1-1: Model of types of genetic susceptibility to disease, ranging from monogenic (Mendelian)

whereby a single rare variant (frequency <0.1%) with a large effect size is sufficient to cause disease, to a

polygenic model whereby multiple common variants (frequency >5%) each with very small effect sizes

contribute to genetic susceptibility to disease. DCM is thought to exhibit monogenic and oligogenic

patterns of inheritance, with the realization that not all familial DCM has an identifiable genetic cause

and that even amongst patients with known variants, penetrance and expressivity are variable. Figure

adapted from(21).

1.3.1.1 Rare genetic variants in DCM

A rare genetic variant alters gene function and occurs at low frequency in a population. In the

25 years following the landmark study linking sarcomeric gene mutations with hypertrophic

cardiomyopathy (HCM) (22), there has been a rapid growth in our understanding of the

genetic basis of cardiomyopathies, including DCM. The genetic architecture of DCM has

emerged to be particularly complex, with almost 70 genetic loci and mutations affecting a

range of diverse cellular structures and functions, most notably with the (Table

1-2). In addition to heterogeneity, there is marked allelic heterogeneity, whereby

different variants in the same gene can cause a similar phenotype. Of note, different variants

29 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

in the same gene can also produce contrasting phenotypes (for example the TNNI3 gene can

cause HCM and DCM).

Rare genetic variants are typically defined as having a minor allele frequency (MAF) <1%,

though the frequency cut offs in the literature vary(23). For evaluation of potentially

pathogenic variants, a disease specific cut off informed by disease prevalence, penetrance,

and allelic contribution to disease is now recommended(24,25).

Table 1-2: Genes implicated in monogenic dilated cardiomyopathy and their cellular component.

Reproduced from Tayal et al (with permission)(26). Genes with the strongest evidence linking them to

DCM are marked with (*) and the most recently identified genes from 2011 onwards are marked with (#).

Causes of predominantly autosomal recessive DCM and older gene associations that have not been

replicated have not been included.

Gene Protein Function Estimated contribution in DCM patients and phenotypic comments Sarcomeric

MYH7* Myosin-7 (beta myosin heavy Muscle contraction Non truncating variants: 5% chain) TNNT2* Cardiac muscle troponin T Muscle contraction Non truncating variants: 3%

TTN*, # Titin Extensible scaffold/Molecular Truncating variants: 15-25% spring TPM1* Tropomysin Muscle contraction <2%

MYBPC3 Myosin-binding protein C, Muscle contraction Major hypertrophic cardiac type cardiomyopathy gene; purported association with DCM now less likely in light of population variation data(14) TNNC1 Cardiac muscle Muscle contraction Mutations also associated with hypertrophic cardiomyopathy TNNI3 Cardiac muscle troponin I Muscle contraction Mutations also associated with hypertrophic cardiomyopathy MYL2# -2 Regulation of myosin ATPase Mutations also associated activity with hypertrophic cardiomyopathy FHOD3# Formin homology 2 domain Sarcomere organization containing 3

DES* Contractile force transduction <1%

30 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Gene Protein Function Estimated contribution in DCM patients and phenotypic comments DMD* Contractile force transduction In patients with dystrophinopathies. X-linked VCL Cell-matrix and cell-cell adhesion

Nuclear Envelope

LMNA* Lamin-A/C Nuclear membrane structure 4%

Mitochondrial

TAZ Tafazzin Associated with syndromic DCM (e.g. Barth syndrome). X-linked Spliceosomal

RBM20 RNA Binding protein 20 Regulates splicing of cardiac 2% genes Sarcoplasmic reticulum

PLN Phospholamban Sarcoplasmic reticulum calcium <1% regulator; inhibits SERCA2a Linked to an pump arrhythmogenic phenotype Desomosomal

DSP* Desmoplakin Desmosomal junction protein Truncating variants: 3% Linked to arrhythmogenic right and left ventricular cardiomyopathy DSC-2# Desmocollin-2 Desmosomal junction protein Linked to arrhythmogenic right and left ventricular cardiomyopathy DSG2# Desmoglein-2 Desmosomal junction protein Linked to arrhythmogenic right and left ventricular cardiomyopathy PKP2# -2 Desmosomal junction protein Linked to arrhythmogenic right and left ventricular cardiomyopathy; recent studies cast doubt on involvement in DCM JUP Junction Desmosomal junction protein Linked to arrhythmogenic right and left ventricular cardiomyopathy Ion channels

SCN5A sodium voltage-gated channel Sodium channel <2%. Associated with atrial alpha subunit 5 arrhythmias and conduction disease. Association with DCM in absence of segregation less strong in light of population variation data(14) Z disc

FLNC# Filamin C Structural integrity of cardiac - myocyte; cross linking protein NEBL Nebulette Z disc protein -

NEXN Nexilin Encodes a filamentous actin - binding protein CSRP3 Cysteine and glycine rich Mechanical stretch sensing - protein 3 TCAP Telethonin Mechanical stretch sensing -

31 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Gene Protein Function Estimated contribution in DCM patients and phenotypic comments LDB3 Lim domain-binding 3 Z disc structural integrity Associated with left ventricular non compaction phenotypes CRYAB Crystallin alpha-B Heat shock protein

Other

BAG3# BAG family molecular Inhibits apoptosis - chaperone regulator 3 ANKRD1 repeat domain 1 Encodes CARP, a transcription <2% coinhibitor RAF1# RAF1 protein MAP3 kinase, part of the Ras- ~9% in childhood-onset MAPK signaling cascade DCM (one study)

Transcription factors

PRDM16# PR domain containing 16 Transcription factor Mutations cause cardiomyopathy in 1p36 deletion syndrome; also linked to isolated DCM and left ventricular non- compaction ZBTB17# Zinc-finger and BTB domain- Transcription factor containing protein 17 TBX5# Cardiac T-box transcription Transcription factor Associated with congenital factor 5 heart disease; also linked to adult onset DCM NKX2-5# NK2 homeobox 5 Transcription factor Associated with congenital heart disease; also linked to adult onset DCM GATA4# GATA Binding Protein 4 Transcription factor Linked to sporadic and familial DCM TBX20# Cardiac T-box transcription Transcription factor Associated with congenital factor 20 heart disease; also linked to adult onset DCM

1.3.1.2 Sequencing methodologies used to detect rare genetic variants

A number of methodologies have been used to understand the genetics of human disease.

Historically candidate gene studies and linkage studies underpinned Mendelian gene

discovery. Linkage studies establish the probability that a given genomic region is associated

with a phenotype, usually in an extended pedigree. These were time consuming, meticulous

studies, requiring large, well characterised families with clearly inherited disease. Traditional

Sanger capillary based sequencing or more contemporary next-generation DNA sequencing

(NGS) approaches such as targeted (panel based) sequencing, whole genome sequencing

(WGS), or whole exome sequencing (WES) are now used to identify rare variants.

32 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

In 1977, Sanger and collaborators published the landmark DNA sequencing technique that

would become and remain the gold standard for DNA analysis(27). Automated Sanger

sequencing continues to be used as a complementary technology to NGS to sequence small

regions, regions difficult to target with NGS (for example those with high GC content) and to

validate NGS findings. The automated Sanger method underpinned the enormous 10 year

collaborative effort to sequence the entire , published in 2001(28). Whilst

whole genome sequencing was then feasible, the cost and timescale prohibited routine use for

clinical or research purposes.

In the era of Sanger sequencing, the study of disease-gene associations commonly involved

sequencing the coding regions of a small number of genes. Pathogenic status was often

ascribed to any rare variant in these genes that was absent in a small control group, typically

~100 samples.

The subsequent advent of next generation DNA sequencing enabled a paradigm shifting

increase in high-throughput sequencing capability, making it possible to sequence large

numbers of genes simultaneously in many patients, thereby decreasing cost and sequencing

time compared to Sanger sequencing.

The term NGS, originally referred to second-generation systems after the ‘first generation’

automated Sanger methodology. Now however, whilst we are in ‘fourth’ generations and

beyond of technology, the term is used more broadly to refer to all high-throughput,

massively parallel sequencing systems.

The fundamental principles common to all NGS strategies is to sequence millions of short-

33 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

DNA fragments in massively parallel arrays, before aligning and mapping the short reads to a

reference genome (overview shown in Figure 1-2). Key methodology consists firstly of

fragmentation of genomic DNA, followed by library preparation, enrichment of target

regions of interest (either by multiplexed PCR methods or in solution oligonucleotide

hybridization methods capturing baits with streptavidin beads), and then massively parallel

sequencing of all captured fragments. The sequencing data is then processed through

bioinformatics pipelines, which involve quality assessment of the raw sequence data, before

alignment of the reads against the reference genome, variant calling, and variant annotation.

Figure 1-2: Overview of NGS workflow

For targeted panel based sequencing, only a predefined set of genes with evidence linking

them to the disease of interest are sequenced. The advantage of this approach is that the depth

of coverage may be greater than for example WES or WGS, more samples can be processed

in a single run, and the bioinformatics approaches to variant calling and annotation, whilst

not trivial, are more straightforward than with WES or WGS. Of course, as all genes are pre-

selected, targeted panel sequencing is not suitable for unbiased gene discovery studies.

With WES, all coding regions of the genome are sequenced. The exome represents ∼1% of

34 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

total genomic DNA and consists of ∼20 000 genes, 180,000 exons, and 30 million base pairs

of DNA. In DCM, WES has been used in patients who are likely to have a genetic basis of

disease but who do not have an identified mutation in known DCM genes. This approach has

led to the discovery of novel DCM variants, for example novel filamin C splicing variants in

3 large families with arrhythmogenic DCM without skeletal muscle defects(29). A key

weakness of WES is the dependence on a PCR amplification step and an exon capture method,

with the capture efficiency being potentially variable for different exons.

In WGS, all 3 billion base pairs of the human genome are sequenced, covering both coding

and non-coding regions. Whilst comprehensive, this is the most costly approach of the 3 NGS

strategies. Like WES, WGS can be used for gene discovery and compared to WES, offers the

advantage of being able to interrogate regulatory regions.

The limitations of WES, and to a greater extent WGS, relate to the difficulty obtaining

adequate sequencing quality, particularly coverage of known disease genes, and in the

informatics challenge of interpretation of the high number of variants discovered, many of

which are unlikely to be of clinical significance.

1.3.1.3 Challenges in the interpretation of rare genetic variants

As outlined, NGS identifies a large number genetic variants per individual. After

bioinformatics processing including variant calling and annotation, variants are then filtered

and interpreted for potential clinical significance. This can be a labour intensive process,

placing the effect of a single variant in the context of existing biological data.

The interpretation of potentially disease causing rare variants is made even more challenging

35 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

due to the relatively high frequency of rare genetic variation in the population. This means

that that an individual variant may be rare (allele frequency <0.001), but collectively,

variation in a specific gene is common. There are a number of population databases such as

1000 Genomes Project (1000GP)(30), the NHLBI GO Exome Sequencing Project (ESP)(31),

the UK 10K Consortium project(32), and the Exome Aggregation Consortium (ExAC) (15).

The ExAC dataset is an aggregation of exome sequencing data from 60,706 individuals from

several large scale studies, including samples from 1000GP and ESP. All raw data from the

various disease-specific and population genetic studies were re-processed and jointly variant

called to enable consistency of bioinformatics analysis across datasets. The samples are not

necessarily from healthy individuals and include patients from the myocardial infarction

genetics consortium (Table 1-3), but are not expected to be enriched for patients with

cardiomyopathy. No individual phenotype information is available, but frequency counts for

any particular variant are broken down for 7 separate ethnic groups (Non-Finnish Europeans,

Finnish, African, Latinos, East Asians, South Asians, Others).

Table 1-3: Composition of the ExAC dataset

Consortium/Cohort Samples 1000 Genomes 1,851 Bulgarian Trios 461 GoT2D 2,502 Inflammatory Bowel Disease 1,675 Myocardial Infarction Genetics Consortium 14,622 NHLBI-GO Exome Sequencing Project (ESP) 3,936 National Institute of Mental Health (NIMH) Controls 364 SIGMA-T2D 3,845 Sequencing in Suomi (SISu) 948 Swedish Schizophrenia & Bipolar Studies 12,119 T2D-GENES 8,980 Schizophrenia Trios from Taiwan 1,505 The Cancer Genome Atlas (TCGA) 7,601 Tourette Syndrome Association International Consortium for Genomics 297 (TSAICG) Total 60,706

36 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Over 7.4 million variants have been identified in gene-coding regions in ExAC(15), with the

majority not entered in ClinVar, a database of disease variants(33). This highlights the need

for use of robust population-level control data to avoid spurious gene-disease associations.

We, and others, have shown how ExAC can be leveraged to aid the interpretation of rare

variants in cardiomyopathies (14,18). However, individuals appear to carry many unique

(private) variants that do not cause disease, with more than 60 unique coding variants found

on average in each participant in ExAC, further complicating approaches to rare variant

interpretation(15).

Therefore population data should also be placed in the context of other available resources to

aid clinicians and researchers in interpreting rare variants, such as disease variant databases

(for example, Human Gene Mutation Database (HGMD)(34), and ClinVar(33)),

computational data (such as in silico missense variant prediction tools, many of which are

amalgamated in the dbNSFP(35)), functional data and, critically, segregation data. Using

these data, variants can usually be classified into four main classes: known pathogenic, likely

(probably) pathogenic, variants of uncertain significance (VUS), and benign variants. The

recent American College of Medical Genetics and Genomics report provides comprehensive

guidelines on variant interpretation (36).

However, conflict may arise between these sources, leading to a greater proportion of

variants being categorised as of uncertain significance instead of likely-pathogenic or

pathogenic, and a degree of subjectivity remains, meaning that even experts can disagree with

respect to variant classification(37,38). Clinical Genome Resource (ClinGen) is a National

Institutes of Health programme to bring together expert panels to agree on approaches to

variant interpretation and curation, to establish clinical validity, pathogenicity and clinical

37 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

usefulness of a given variant in a specific disease(39). Whilst such approaches are admirable,

they are resource and personnel intensive, and may not be able to keep pace with the rapid

discovery of novel disease-gene associations and potentially novel variants within established

genes.

1.3.2 Disease-gene associations in DCM

Whilst variants in almost 70 genes have been linked to DCM (Table 1-4), the evidence is

most robust for a “core disease set” encompassing the sarcomeric genes MYH7 (which

encodes beta myosin heavy chain), TNNT2 (which encodes troponin T2), TTN (encoding

titin) and the nuclear envelope protein encoding gene LMNA. Many of the existing predictive

tools used to distinguish benign variants from causative mutations are of limited utility in the

NGS and population database era, casting doubt on earlier findings from candidate gene

sequencing, where unclassified variants were deemed pathogenic based on their absence in

small control cohorts or limited in vitro data(40,41).

38 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Table 1-4: Genes linked to DCM, annotated by year of report in HGMD

Year Genes Year Genes Year Genes 1996 DMD 2006 CRYAB DOLK 1998 ACTC1 FKTN JUP 1999 DES DNAJC19 PKP2 LMNA FLT1 BAG3 2000 CTF1 PSEN2 TXNRD2 DSP SYNM 2013 ISL1 MYH7 PSEN1 EMD SGCD 2007 FOXD4 LAMP2 TNNT2 PDLIM3 PRDM16 2001 TPM1 ILK NPPA 2002 TTN LAMA4 SGCB CSRP3 2008 MYPN GATA4 MYBPC3 CHRM2 FHOD3 TCAP DSG2 ACTA1 VCL LAMA2 2014 GATA6 2003 ACTN2 2009 ANKRD1 CASQ2 LDB3 RBM20 RAF1 PLN NEXN GATA5 2004 ABCC9 2010 TAZ NKX2-5 SCN5A NEBL 2015 ZBTB17 TNNC1 SYNE1 TNNI3 TCF21 2005 MYH6 2011 GATAD1 TMPO MURC DSC2

A recent large scale analysis of rare genetic variation in cardiomyopathy cases from our

group compared to population variation has provided insights into the genetics of DCM. The

study tested for an excess of rare variants in 46 genes sequenced in up to 1315 DCM cases

compared to over 60,000 ExAC reference samples. Truncating variants in TTN were the most

common DCM rare variant (14.6%)(14). There was modest, statistically significant

enrichment in only 6 other genes (MYH7, LMNA, TNNT2, TPM1, DSP and TCAP) (Table

1-2). Based on available data, RBM20 is also likely to prove to be a significant disease gene

(reviewed below) but was not included in the published analysis due to poor coverage in the

ExAC data. Furthermore, sequencing methods were not uniform, and not all genes were

sequenced across the DCM cohorts included in the study. Even allowing for this however,

many genes that have previously been linked to DCM, including genes routinely sequenced

39 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

in clinical practice such as MYBPC3 and MYH6 showed little or no excess burden in DCM

compared to the reference population. The accompanying Atlas of Cardiac Genetic Variation

web resource (14) summarizes this data and serves as a useful adjunct to facilitate the

interpretation of rare variants in DCM.

In the following section, I now briefly review the notable established DCM genes, before

reviewing genes recently reported to be associated with DCM. Titin will be discussed more

comprehensively in Section 1.4.

1.3.2.1 Established disease-gene associations in DCM

Sarcomeric

Mutations in the sarcomeric proteins responsible for the generation and regulation of cardiac

contraction can lead to DCM, as well as HCM. For DCM, these mutations appear to lead to

reduced calcium sensitivity(42).

Mutations in proteins of both the thick and thin cause DCM. The thin filament

protein cardiac actin (ACTC1) was the first sarcomeric gene linked to DCM but mutations are

very rare(43). Other sarcomeric mutations in DCM were originally identified after screening

of the genes MYH7, MYBPC3, TNNT2 and TPM1 in 46 patients with DCM(44).

Most commonly, β-Myosin heavy chain (MYH7) mutations are found in 5% DCM, though

with marked allelic heterogeneity(44). β-Myosin heavy chain plays a key role in force

generation by myosin. In the original study, two MYH7 mutations (Ala223Thr and

Ser642Leu) were found in two patients(44). Ser642Leu was part of the actin-myosin interface

40 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

and Ala223Thr affected a buried residue near the ATP binding site.

Sarcomeric contraction is regulated by tropomysin and the troponin complex ( I, T,

and C). Mutations in the gene encoding Troponin T (TNNT2), which binds to tropomysin, are

found in <5% of DCM(42,45).

Other sarcomeric genes identified in familial DCM are α- (TPM1), troponin C

(TNNC1), and troponin I (TNNI3). Troponin I inhibits actin-myosin binding during diastole

and Troponin C (TNNC1) binds calcium during systole to promote cross bridge formation

between actin and myosin, playing important roles in the regulation of contraction(42,46).

The thick filament core is formed by β-myosin heavy chain, with myosin binding protein C

(MYBPC3) highly concentrated in the M band region. Mutations in MYBPC3 have previously

been linked to DCM, with the original sarcomeric gene study identifying one missense

mutation in one individual out of 46 DCM patients, absent from 88 controls(44). However, it

has now been established that this gene has a high background level of variation and may not

be a causative DCM gene(14).

Nuclear proteins

The lamin A/C gene (LMNA) encodes the nuclear envelope proteins lamin A and lamin C. In

addition to DCM, mutations in LMNA cause a diverse range of phenotypes including Emery-

Dreifuss muscular dystrophy, limb-girdle muscular dystrophy, lipodystrophy, progeria and

restrictive dermopathy. LMNA missense variants were first linked to DCM after evaluation of

11 families with autosomal dominant DCM and conduction disease when 5 novel missense

variants were identified(47). The original and subsequent studies have demonstrated that

LMNA associated DCM is nearly always preceded by conduction disease and atrial and

41 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

ventricular arrhythmias, and is often associated with skeletal involvement, early onset

cardiomyopathy and a higher risk of sudden cardiac death(47-49). LMNA associated DCM is

also highly penetrant, with manifestation of the phenotype in mutation carriers by 50-60

years of age. LMNA mutations have also been associated with an LV non compaction

phenotype(50).

1.3.2.2 Recent disease-gene associations in DCM

Since 2005, 47 new genes have been categorized as linked with DCM in the HGMD(34)

(Table 1-4). Many of these links have not been replicated outside of the original reports.

Examples of the most notable novel associations are discussed below, selected for critical

evaluation either due to robust evidence, novelty, or clinical importance.

BAG3

BAG3 encodes a heat shock chaperone protein and was first linked to DCM in 2011 through

the discovery of a large 8733bp deletion in exon 4 in 7 affected family members in a three

generation family, which was absent in 355 controls (51). Subsequently, coding exons in

BAG3 in 311 other unrelated DCM probands were sequenced, which identified 7 rare variants

(1 frameshift, 2 nonsense, and 4 missense variants) that were absent from 355 controls. The

authors were also able to recapitulate the DCM phenotype in a zebrafish bag3 knockdown

model. In separate studies, BAG3 was linked to DCM through a genome wide association

study (GWAS), with the discovery of a non-synonymous SNP in the coding sequence of

BAG3 in DCM cases compared to healthy controls (rs2234962, p=1.1x10-13) (52). The

authors then performed targeted sequencing in a cohort of 168 unrelated DCM probands and

identified 6 variants that were also detected in affected relatives, lending further support to

the role of BAG3 as a disease causing gene.

42 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

RBM20

RBM20 encodes a spliceosome protein that regulates pre-mRNA splicing for many genes,

including TTN (53), which is why variants in this gene may hold particular relevance for

DCM, either in isolation or in digenic heterozygosity with TTN(54). RBM20 was initially

associated with DCM through linkage analysis in 2 large families with DCM (55). The

authors sequenced all 14 RBM20 exons in each family member and identified a heterozygous

missense mutation in exon 9 that co-segregated with disease in all affected individuals, and

that was absent in unaffected relations and 480 ethnically matched controls. The authors went

on to detect RBM20 missense mutations in exon 9 in 6 more families affected with DCM.

Since the original link with DCM (55), subsequent studies found mutations both within and

outside the original RBM20 hotspot in DCM probands, but the segregation data on these

variants is limited and the control population was modest in size, meaning that population

level missense variation was not accounted for in these regions (56,57). The association of

RBM20 and DCM appears most robust for variants in the original hotspot and further

curation is needed to understand the significance of variants in other regions.

PRDM16

The 1p36 deletion syndrome can be associated with cardiomyopathy, and the PRDM16 gene

(which encodes a transcription factor) was identified as a possible cardiomyopathy gene at

this locus, linked with a syndromic cardiomyopathy as well as with adult onset DCM (in 5

out of 131 individuals with 4 novel missense variants) (58). However, whilst there may be a

role for PRDM16 in cardiac development, its role as a cardiomyopathy gene has subsequently

been questioned (59).

ZBTB17

ZBTB17 is also encoded on 1, at the 1p36 locus. A study of cardiac myocytes

43 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

and a mouse model of ZBTB17 deletion demonstrated that ZBTB17 is involved in cardiac

myocyte hypertrophy and is essential for cell survival (60). The authors also identified that

ZBTB17 encodes a transcription factor (zinc-finger and BTB domain-containing protein 17)

that binds the gene CSRP3, a Z disc protein, mutations of which are found in both HCM and

DCM. Given the association between CSRP3 and DCM (in a small cohort with limited

segregation data (61), with no subsequent replication), and this new found function of

ZBTB17 in binding CSRP3, the authors hypothesized that ZBTB17 could be a novel gene

implicated in DCM.

PLN

Phospholamban is a regulatory protein that controls calcium cycling in cardiomyocytes

through inhibition of the cardiac SERCA isoform leading to reduced calcium influx in the

sarcoplasmic reticulum. Mutations in the gene encoding phospholamban (PLN) were first

linked to DCM in 2003(62)and subsequently to heart failure through the study of a large

pedigree with hereditary heart failure(63). In the latter study, a heterozygous deletion of

arginine 14 in the coding region of the gene (PLN-R14Del) was found in all affected

individuals who had a phenotype of LV dilation, contractile impairment and ventricular

arrhythmia. The human phenotype was replicated in transgenic mice.

The PLN-R14Del has since been detected in larger cohorts of patients with DCM, although in

one study it was associated with a mild DCM phenotype(64), in contrast to the severe

phenotype in the original publication. The prevalence of the PLN mutations was higher in

Dutch studies (up to 15% of DCM)(65,66) compared to other cohorts (<1% of DCM)(67).

The PLN-R14Del has now been shown to be a founder mutation in the Netherlands(68).

44 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Transcription factors

Many additional transcription factors have also been linked to DCM in recent years, such as

GATA5(69), TBX20(70), TBX5(71), GATA6(72), GATA4(73) and NKX2-5(74). Some of these

genes are clearly linked to congenital heart disease phenotypes. However, many of the

variants claimed to be associated with DCM are missense variants that have been identified

within one relatively small group of DCM patients, with variable segregation data. Further

studies are required to confirm the link with DCM.

Desmosomal proteins

Desmosomal proteins, typically perturbed in arrhythmogenic right ventricular

dysplasia/cardiomyopathy (ARVD/ARVC) have also been linked to DCM. The association

has been most robust for DSP which encodes desmoplakin, a desmosomal protein (75), with

a strong excess of truncating variants in DSP in DCM(14). However, some of the more recent

associations of desmosomal protein gene variants have limited variant curation and

segregation data, such as PKP2 (76) (which encodes plakophilin 2), and these associations

are less clear. One such PKP2 variant (c.419C > T(p.(S140F)), previously linked to DCM has

been shown not to be associated with heart failure phenotypes (77). Therefore, of the

desmosomal proteins, DSP variants have the most robust association with DCM.

Filamin C

Filamin C (encoded by FLNC) is a Z disc protein that provides sarcomeric stability. In recent

work, two rare splicing variants in FLNC were detected though whole exome sequencing in

two Italian families and in one US family affected with DCM, with all variants co-

segregating with disease (29). Only one unaffected variant carrier was identified but this

individual declined further follow up. These variants were absent from 1000 Genomes,

NHLBI Go-ESP and ExAC. The FLNC cardiomyopathy phenotype was not associated with

45 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

skeletal muscle involvement in this cohort, but was associated with arrhythmias and sudden

cardiac death. In the same study, a zebrafish knockdown model showed a phenotype of

cardiac dysfunction, with defects in the Z discs and sarcomere disorganization. Evaluation of

FLNC variants in a large (n=2,877) cohort of patients with inherited cardiac diseases,

including DCM, has shown that the phenotype of individuals with truncating variants in

FLNC is notable for left ventricular dilation, systolic impairment, ventricular arrhythmias,

cardiac fibrosis and sudden cardiac death (78). Further replication in DCM-specific cohorts is

needed to validate this potentially prognostically important phenotypic association.

Summary of recent disease-gene associations in DCM

In summary, there have been many novel gene and variant associations with DCM. Whilst

some appear robust and potentially clinically important (e.g. FLNC, BAG3, RBM20), others

require further study (e.g. variants in transcription factors). As yet, there is insufficient clarity

on the importance of variants outside of major disease genes and this remains a major unmet

need to be addressed. As the commonest genetic cause of DCM identified to date, truncating

variants in the titin gene (TTNtv) are now reviewed separately to the other genes.

46 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.4 Truncating variants in titin in DCM

1.4.1 Function of the titin protein

The TTN gene (363 exons) encodes the largest-known human protein, titin (~35,000 amino

acids, ~2µm long). One molecule spans half the sarcomere from the Z disk to the M line. Its

main roles are in the contractile process, regulating sarcomeric contraction, stretch sensing

and signalling, interacting with numerous cellular proteins(5,79). Titin is the main source of

cardiomyocyte resting tension and stiffness(80).

The titin protein is composed of 4 modular domains, corresponding to their location in the

sarcomere(80,81). At the N-terminus, titin is embedded in the Z-disk of the sarcomere and at

the C-terminus, it is embedded in the M-band region. The core of the protein is composed of

the A and I bands. The A band is inextensible and is anchored to the thick filament. It plays a

role in active force generation and length dependent activation of the sarcomere. The highly

extensible I band is composed of repeat tandem immunoglobulin segments, a spring like

PEVK domain, and N2A and N2B elements. This section is responsible for the passive

elasticity of muscle. During active contraction, titin is extended in the opposite direction to

that which occurs in passive stretch, which results in a cardiomyocyte restoring force and

elastic recoil on relaxation. This property may contribute to ventricular diastolic suction(81).

The mature human heart has two main titin isoforms formed from alternative splicing of large

fetal type isoforms; the larger compliant N2BA isoform (containing the N2A and N2B

segments) and the shorter stiffer N2B form, differing in the length of the immunoglobulin

and PEVK repeat segments(81). The ratio of both isoforms is approximately 40:60

N2BA:N2B in adult myocardium(80,82), but the ratio switches in response to physiological

47 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

perturbations and can occur in heart failure(81). As well as isoform switching, the stiffness of

titin is modulated by post-translational modifications, including phosphorylation by cyclic-

AMP–dependent protein kinase (PKA) and cyclic-GMP–dependent protein kinase (PKG)

(both decreasing stiffness) and protein kinase C alpha (increasing titin stiffness)(80,82).

1.4.2 Prevalence of titin truncating mutations in DCM

Variants in titin were first associated with DCM in 2002 through the study of two large

multigenerational families affected with DCM(83). In the first kindred, linkage analysis

identified a disease gene locus (maximum logarithm of odds (LOD) score 5.0, penetrance of

70%). In this study, TTN was chosen as a candidate gene due to high levels of cardiac

expression and its established role in muscle assembly and function. A 2bp insertion was

identified in exon 326, that resulted in a frameshift mutation generating a premature stop

codon, and this mutation segregated with disease in family members. In the second kindred, a

non-truncating TTN missense mutation in a highly conserved region was identified that also

segregated with disease (Trp930Arg).

More recently, next generation sequencing technologies have made the study of the giant titin

gene possible in large cohorts. This led to the discovery that truncating variants in TTN are

found in approximately 15% of unselected DCM cases and in up to 25% of end-stage DCM

cases (79,84).

1.4.3 Established genotype phenotype correlations in TTNtv DCM

There have been 4 key studies evaluating genotype-phenotype correlations to date in TTNtv

48 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

DCM. They are summarized in Table 1-5. Establishing clear genotype-phenotype

correlations in TTNtv DCM is of interest because of the number of DCM patients affected

with TTNtv, as the commonest genetic contributor to DCM. There is therefore the potential

to improve patient stratification for a large proportion of DCM patients.

In the first notable cohort based phenotype study of TTNtv, there were 67 patients with

TTNtv amongst 312 patients with DCM from 3 distinct cohorts(79). There were no distinct

phenotypic differences at baseline between TTNtv positive and negative patients with respect

to LVEF or LV end diastolic diameter. The authors demonstrated earlier onset of adverse

events (cardiovascular death, left ventricular assist device implantation, and heart transplant)

in male patients with TTNtv compared to female patients. However, the 3 cohorts were fairly

distinct, including a relatively mild familial cohort and an end stage cohort, which may have

affected the ability to discriminate phenotypes or outcome differences. All patients in the

study underwent echocardiography based phenotyping and there was no reporting of

reproducibility in measurements between centres, particularly important as echocardiography

is known to be operator dependent, which may have confounded genotype-phenotype

correlation. Finally, the comparison of age at onset of adverse events between males and

females was only made within the TTNtv positive group and not in comparison to the overall

DCM cohort. The finding therefore could simply reflect the combination of a male

preponderance in DCM and an adverse prognosis with DCM overall.

In the next study, from our centre, 42 patients with TTNtv from a cohort of 319 patients with

DCM were evaluated(84). Whilst single centre, all patients underwent CMR phenotyping.

There was no difference in age of onset of disease between TTNtv positive and negative

groups. In this pilot analysis, the authors demonstrated a borderline reduction in LVEF and

49 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

right ventricular ejection fraction (RVEF) in TTNtv positive patients compared to TTNtv

negative patients. They also showed evidence of a length dependent positional effect in

severity of disease, with more distal truncations demonstrating a more severe cardiac

phenotype. Although the sample size was relatively small, in outcome analysis, they

demonstrated a trend towards reduced survival in TTNtv positive patients compared to

TTNtv negative patients. Gender specific differences in outcome were not reviewed.

The next TTNtv phenotype study consisted of 118 patients with TTNtv, to be commended as

the largest study of TTNtv DCM patients to date. However, this cohort only included 45

probands. The remaining 73 individuals with TTNtv were detected through family

screening(85). The cohort was also subject to ascertainment bias, as the overall DCM cohort

was only 235 patients. It was not representative of an unselected DCM cohort. The remaining

non-TTNtv patients included 57 patients with LMNA variants and 60 patients who were

mutation negative. Therefore the phenotypic comparisons were largely driven by a

comparison of TTNtv DCM versus LMNA DCM. Allowing for this, there was no difference

in age, gender, or the presence of conduction disease between TTNtv positive and negative

patients, but there was a borderline reduction in LVEF in TTNtv positive patients. Patients

were followed up for a median of 2.5 years for the composite endpoint of all cause mortality,

sudden cardiac death, resuscitation, ICD implantation, heart transplantation or left ventricular

assist device implantation. It appeared that TTNtv were associated with better outcomes than

non-TTNtv DCM. However, this finding was largely driven by the high adverse event rate

known to be associated with LMNA cardiomyopathy. Therefore the authors were able to

demonstrate that TTNtv DCM did not have the adverse prognosis known to occur with

LMNA cardiomyopathy, but they did not provide sufficient evidence to demonstrate that

TTNtv DCM was associated with a more favourable prognosis compared to idiopathic

50 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

mutation negative DCM.

Subsequently, the comparison between TTNtv and LMNA cardiomyopathy was contradicted

in a Finnish study of 145 patients with DCM, including 25 patients with TTNtv and 12

patients with LMNA variants. There was no difference in the rate of the composite end point

of cardiac transplant or death between TTNtv patients and patients with LMNA variants(75).

Whilst this was a much smaller study, follow up time was longer (mean 6 years). However,

the analysis was limited by studying age of onset of adverse event, as opposed to onset of

event as per duration of follow up. Such analysis strategy is limited because the majority of

patients in the study were aged between 40-60 years old, and it may be too early to detect

adverse events in adult onset DCM.

The most recent phenotype study of TTNtv consisted of 46 patients with TTNtv, of which 17

were TTNtv DCM probands, with 29 mutation carrier relatives, of whom only 9 were

definitely affected with DCM(86). The overall DCM cohort was 72 patients. There was no

difference in the age, gender, or cardiac phenotype between TTNtv positive and negative

individuals. On mean follow up of 63 months for the composite end point of heart failure

death, heart transplant, or left ventricular assist device implantation, there was no difference

in the rate of adverse events between TTNtv positive and negative individuals. Interestingly,

the authors reported worse outcomes in males with TTNtv (n=24) compared to females with

TTNtv (n=21), highlighting a possible female gender protective effect in the presence of

TTNtv. This had been alluded to in previous studies but this was the first study to explicitly

highlight this. However, this analysis was limited because the gender outcome analysis was

performed in all individuals who carried TTNtv variants, and was not limited to those who

were phenotype positive. Therefore the adverse events could have been driven simply by

51 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

those individuals who had developed overt DCM, and not reflect gender specific effects of

TTNtv. The sample size in each group was small and these findings require confirmation in

larger prospective studies evaluating broader cardiovascular end points.

In summary, the phenotype studies of TTNtv DCM to date have differed in the presence and

extent of phenotypic differences in patients with and without TTNtv and the clinical course

of TTNtv DCM, particularly with regards to gender based differences in outcome(79,84-86).

These discrepancies may relate to variations in TTNtv curation between studies, for example

with respect to exon location, percentage spliced in (PSI) score, or variant frequency. They

may also be due to differing patient factors, such as baseline severity of disease at the time of

study recruitment, and differing environmental risk factors. The work in this PhD seeks to

bridge the gap between the previous studies by studying the effects of TTNtv including

clinical outcomes in patients with DCM who have been well phenotyped, with consistent

TTNtv variant curation, together with a comprehensive understanding of background genetic

variation and environmental risk factors.

52 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Table 1-5: Summary of key genotype-phenotype studies of TTNtv DCM. LVAD= left ventricular assist device, OCTx = orthotopic cardiac transplant, LVEDD =

left ventricular end diastolic diameter, LVEF= left ventricular ejection fraction

Study Cohort TTNtv Phenotype Follow Endpoint Outcome Limitation up Herman et al, 312 DCM 67 No difference in age, - Cardiac death, Younger age of onset of Variable phenotype between cohorts, including 2012(79) patients LVEDD or LVEF LVAD, OCTx events in men end stage DCM. Echo based phenotyping – from 3 limited reproducibility between centres. cohorts Outcome gender comparison made within TTN+ only, not compared to DCM overall. Roberts et al, 319 DCM 42 No difference in age. Censored Death, OCTx, A trend to reduced Well phenotyped cohort but relatively small 2015(84) with CMR Borderline reduction at 70 LVAD survival in TTNtv sample size and phenotypic differences (subset of in LVEF and reduced years old. positive patients main paper) RVEF in TTNtv + compared to TTNtv patients. Length negative (p=0.05). dependent positional phenotypic effect. Jansweijer et 235 DCM 118 TTNtv (45 No difference in age, 2.5 years Composite of ICD TTNtv appeared to be Ascertainment bias, cohort selected for TTNtv al, 2016(85) patients probands, 73 gender, conduction (median) implantation, associated with better and LMNA variants, small control cohort size (60 screened disease between TTN+ resuscitation, outcomes but this limiting ability to draw TTN+ vs idiopathic idiopathic relatives) and TTN- patients. sudden cardiac comparison was largely DCM conclusions DCM Borderline difference death, OCTx, driven by the adverse mutation in LVEF. LVAD, all cause prognosis of LMNA negative, mortality cardiomyopathy. No 57 LMNA gender differences in variants, outcome. 118 TTNtv) Franaszczyk 72 DCM 46 TTNtv (17 No difference in age, 63 Heart failure death, No difference in adverse Gender outcome analysis included all TTNtv et al, 2017(86) patients TTNtv gender, or phenotype. months OCTx, LVAD cardiac events between carriers, not limited to those phenotype positive. probands, 29 (mean) probands with and mutation without TTNtv. carrier relatives Reported worse of whom 9 outcomes in males with definitely TTNtv (n=24) compared affected). to females (n=21).

53 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.4.4 Challenges in the interpretation of TTNtv

There are challenges in the interpretation of TTNtv. The finding of TTNtv in apparently

healthy control individuals (estimated at ~3% in early studies(79), reduced to ~0.5% in more

recent studies accounting for exon usage and variant location(84,87,88)), has led to doubt

over whether TTNtv by themselves are truly pathogenic. However, recent work has shown

that on deep phenotyping with 2d and 3d-CMR, TTNtv are penetrant in apparently healthy

individuals, with subtle abnormalities in cardiac structure and function(88).

Like many genetic variants in DCM, TTNtv exhibit variable penetrance and expressivity.

Variable penetrance means that not all individuals who carry TTNtv will develop a

phenotype. Variable expressivity refers to variation in the severity of the resulting phenotype.

This raises the possibility that additional modifiers, either genetic or environmental, are

needed to develop the phenotype of TTNtv DCM.

1.4.5 Evidence for genetic and environmental modifiers of TTNtv

cardiomyopathy

There is a growing body of evidence for the importance of genetic and environmental

modifiers of TTNtv cardiomyopathy. In this model, the presence of a TTNtv in isolation is

not sufficient to cause the DCM phenotype but requires the presence of an additional

environmental or genetic factor (Figure 1-3).

54 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Figure 1-3: Putative model of genetic and environmental modifiers of TTNtv

Genetic modifiers

In principle, genetic modifiers may act at local or distant sites. With local sites, DNA variants

near to the mutated allele or the wild type allele may influence regulation of the gene. In

distant-acting mechanisms, DNA variants away from the mutated allele exert effects, for

example mutations in separate loci.

The finding that digenic heterozygous missense variants in TTN and LMNA were associated

with severe DCM phenotypes(89) showed the potential for modifier genes or additive genetic

effects in DCM. This concept was alluded to in a multi-centre study of 639 patients with

sporadic or familial DCM, with the finding of a 38% rate of compound mutations, and up to

44% when considering patients with TTNtv (19). However, these findings must be

interpreted with great caution as the “yield” of DCM variants in this study was far higher than

in any previous study, background population variation was not well accounted for, and there

were no matched controls on the same sequencing platform. The interaction between TTN

and variants in RBM20 will be an area of interest for future studies given the known role of

RBM20 as a regulator of alternative splicing of TTN(53,90).

Epigenetics refers to a change in that is not explained by DNA sequence

variants, but occurs as a result of alterations to the packaging or translation of genetic

55 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

information(91). These mechanisms can be inherited or acquired and involve 3 key processes

to either activate or silence a gene, namely methylation of CpG islands, modification of

histone proteins, or the role of microRNAs. MicroRNAs are small non-coding RNA

molecules that modify the expression of messenger RNAs by interfering with their

translation. Epigenetic modifiers of disease have not yet been shown in TTNtv

cardiomyopathy but may be a promising area of future research.

Environmental modifiers

With regards to environmental modifiers of TTNtv DCM, the discovery that peripartum

cardiomyopathy shares a genetic aetiology with DCM suggests that pregnancy may act as an

environmental modifier to unmask the phenotype of TTNtv cardiomyopathy (92). More

recently, a case report of 2 patients with chemotherapy induced DCM found to have TTNtv

was published(93), again supporting the ‘second hit’ concept of TTNtv cardiomyopathy.

Animal studies have consistently demonstrated that rodents with TTNtv only developed

impaired cardiac physiology under cardiac stress (88,94), providing further evidence of the

importance of gene-environment interactions in the development of the TTNtv

cardiomyopathy.

Gene-environmental interactions in DCM have been shown to have prognostic importance in

a study of 213 DCM patients(95). In isolation, the presence of genetic DCM did not confer an

adverse prognosis, but when combined with a history of an environmental risk factor such as

a significant viral load, immune factors, arrhythmia or toxic triggers, a worse outcome was

noted(95). Whilst this study did not specifically report on the interaction between TTNtv and

environmental factors, it is plausible that interactions with TTNtv (as the commonest single

56 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

gene contributor to DCM) underpinned some of the findings.

There has been recent interest in the effects of alcohol on the cardiovascular system(96) and

it has long been established that excessive alcohol consumption can lead to

cardiomyopathy(97,98). It has also been recently shown that moderate alcohol consumption,

in excess of UK government guidelines, is associated with an increased risk of heart failure

(hazard ratio 1.22)(99). Not everyone with a history of excess consumption however develops

cardiac complications(98), therefore there may be additional risk factors, such as a genetic

susceptibility(100,101). Against this background, the next logical step would be to evaluate

the interaction between TTNtv and alcohol consumption. To date, it has not yet been

established whether alcohol is a phenotypic modifier of titin cardiomyopathy. The work in

this PhD seeks to address this gap.

This PhD is an integrated study of the genetic and phenotypic assessment in DCM. Having

reviewed the genetic basis of DCM, I now move on to review the role of cardiovascular

magnetic resonance in the phenotyping of DCM.

57 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.5 Role of Cardiovascular Magnetic Resonance in Dilated

Cardiomyopathy

1.5.1 Overview of cardiovascular magnetic resonance

Cardiovascular magnetic resonance (CMR) is a non-invasive imaging technique that does not

use ionising radiation(102). Detailed tissue characterisation and longitudinal reliability and

reproducibility of key measurements make CMR a robust imaging biomarker.

CMR generates images using a combination of radiofrequency pulses in a magnetic field.

Hydrogen-1 nuclei are abundant in water and fat and possess an inherent property of spin.

When placed in an external magnetic field, these nuclei line up in the direction of the

magnetic field. Application of an external radiofrequency pulse causes the direction of the

nuclei to shift and creates local tissue magnetisation. As the magnetisation decays and the

nuclei return to their original state, that is relax, they emit their own radiofrequency signal.

This signal is captured and through complex Fourier transformation, an image of spatially

resolved radio signals is generated. This relaxation can occur in transverse (T2) and

longitudinal (T1) planes. In MR imaging, the signal from any tissue is determined by the

density of hydrogen atoms (proton density), and by the longitudinal relaxation time (T1) and

transverse relaxation time (T2). As the magnetic resonance characteristics of protons varies

between tissues, decoding of these separate signals permits tissue characterisation(103). In

addition, image contrast can be modified by altering the radiofrequency pulses. For example,

in T1 weighted images, myocardial tissue is dark whereas fat is bright, in contrast to T2

weighted images in which water is bright, for example myocardial oedema. A combination of

radiofrequency pulses, magnetic gradient field switches and timed data acquisitions is known

as an MR sequence.

58 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

There are many types of MR sequences, but common categories include spin echo and

gradient echo. Dark blood spin echo sequences are mainly used for anatomic imaging and

tissue characterisation. White blood gradient echo sequences are used for cine images.

Steady-state free precession MRI (SSFP) is the most commonly used sequence to assess

cardiac function, providing excellent contrast between blood and myocardium. It is a type of

gradient echo MRI pulse sequence in which a steady, residual transverse magnetisation is

maintained between successive cycles.

Unlike static structure imaging (e.g. soft tissue or brain), CMR images a moving structure

which requires ECG gating, so that the final image is generated by capturing data from

certain parts of the cardiac cycle and sampling across multiple cardiac cycles. In addition,

respiratory motion artifact is avoided by acquiring CMR images in end expiratory breath

holds, though the success of this depends on patient compliance.

Overall, there are no known detrimental biological side effects of MR imaging if safety

guidelines are adhered to, particularly with regards to application of pulse sequences(102).

There have been recent reports demonstrating immediate post CMR double strand breaks in

DNA in T lymphocytes(104,105), but what the mechanism is, whether these persist, or are

associated with cancer or downstream biological effects is far from established(106).

1.5.2 The role of CMR in DCM

In DCM, imaging is used to establish the diagnosis and assess response to treatment in

probands, as well as screen family members. Imaging findings can also be used to predict

outcomes. CMR is an excellent imaging modality in DCM, particularly in patients with

suboptimal echocardiographic images. As the gold standard for assessment of cardiac

59 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

volumes, combined with being non-invasive and not exposing subjects to ionising radiation,

it is the ideal imaging modality for accurate serial evaluation of patients with DCM. A

particular strength of CMR in DCM is to highlight the presence of additional comorbidity

such as coronary artery disease(107,108) or provide clues as to the underlying aetiology, for

example abnormally short T2* relaxation times in hemochromatosis(109).

This work in this study harnesses the power of CMR in a large cohort of DCM patients and

takes advantage of some of the applications highlighted in the following sections.

1.5.2.1 Imaging biomarkers of DCM identified through CMR

1.5.2.1.1 Morphology and function

DCM is defined by the phenotype of left ventricular chamber dilation and impaired systolic

function. In contrast to echocardiography, CMR can image in any plane with an unrestricted

field of view, allowing evaluation of cardiac structures, regardless of acoustic windows or

body habitus.

It is well established that LVEF is a robust prognostic marker in DCM(110) and forms the

basis for selection of patients for device therapy (111,112). Right ventricular dysfunction,

found in 30-60% of DCM cases, is also an independent predictor of transplant free survival,

adverse heart failure outcomes and arrhythmic events(113,114). Therefore precise

measurement of both is crucial.

CMR is the gold standard non invasive technique for accurate and reproducible assessment of

LV volumes, mass and ejection fraction, crucially without the need for geometrical

60 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

assumptions that are required in 2d echocardiography for example (115,116). Measurements

are obtained through manual planimetry or the use of semi-automated software.

CMR also provides the gold standard non-invasive assessment of RV size and function due to

its 3-dimensional capabilities (117,118). Accurate assessment of RV size and function can be

challenging with echocardiography, due to its complex and variable shape. In addition, the

low inter- and intra-observer variability are great strengths of the use of CMR in DCM,

permitting the early detection of even subtle abnormalities in ventricular function(103,119-

121).

Situations in which CMR may not provide such accurate assessment of cardiac volumes and

function include the presence of motion or breathing artifacts, particularly in patients too

symptomatic to complete breath holding or in patients with irregular rhythms, where high

quality cine imaging may not be acquired.

CMR also allows accurate quantification of left atrial (LA) volume using the biplane area-

length method(122). This compares favourably against other non-invasive imaging methods

due to its excellent endocardial border definition and multiplanar imaging ability, even in the

presence of atrial fibrillation (123-126). LA size is often increased in cases of DCM

secondary to pressure overload from LV diastolic impairment, functional mitral regurgitation

and atrial fibrillation. It has been proposed that the degree of LA enlargement reflects the

degree of diastolic dysfunction and LA size has been used to predict heart failure outcomes

(125). It has also been demonstrated that LA volume indexed to body surface area calculated

using CMR independently predicts cardiac transplant-free survival in DCM, with a cut-off

value of >72mL/m2 predicting a three-fold increase in adverse outcomes (127).

61 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Functional mitral regurgitation is also a common consequence of DCM secondary to mitral

annular dilatation and leaflet tethering secondary to left ventricular impairment. A long-axis

cine stack and a short axis cine image across the mitral valve allows accurate assessment of

all parts of the valve apparatus including individual leaflet scallops, chordae and papillary

muscles (128). Mitral regurgitant volume can be calculated by estimating the aortic forward

flow volume, using phase contrast flow imaging, and subtracting this from the total left

ventricular stroke volume(128,129). This method has been validated against volumes

calculated from echocardiographic indices and catheterisation data with good inter-technique

agreement (129,130). Once again, the accurate assessment of the degree of functional mitral

regurgitation provides important prognostic information in DCM (131).

Given the accuracy in functional assessment, CMR has been proposed as the method of

choice for serial evaluation of patients with DCM after therapeutic intervention (132,133).

Given the favourable interobserver variability compared to other methods of assessment, the

use of CMR in clinical trials can reduce the sample size required, reducing the overall cost

and time needed to complete the research (134).

1.5.2.1.2 Tissue characterisation

The importance of tissue characterisation

In addition to its strengths in quantification of cardiac dimensions and function, CMR also

excels in non invasive tissue characterisation. Microscopically, cardiomyocytes in DCM are

characterized by myocyte hypertrophy, myocyte loss, and interstitial fibrosis(135). Tissue

characterisation in DCM holds the promise to characterise subphenotypes of disease which

62 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

may offer insights into disease pathogenesis, identify novel therapeutic targets, monitor

disease progression or facilitate prognostication.

In particular, leveraging the detailed in vivo tissue phenotyping that advanced CMR offers

may enable precise phenotypic correlations in genetic cardiomyopathies(136). Current

imaging biomarkers do not permit the detection of specific imaging signatures of genetic

cardiomyopathies(137). At present, it is unknown whether this is because genetic variants are

not associated with a specific phenotype or whether subtle/latent changes are not properly

ascertained.

Non-contrast tissue characterisation

Myocardial tissue can be characterized through modified pulse sequences. For example, the

T2 weighted Short-Tau Inversion Recovery (STIR) sequence can image oedema, suppressing

signal from fat and flowing blood. Long T2 relaxation times of protons in water generate high

signal in areas of myocardial oedema(138). The sequence was initially validated in ischaemic

models, but is now used in cardiomyopathies, particularly to differentiate acute from chronic

myocarditis(139). The interpretation of STIR images however is subjective and novel direct

quantification of T2 by mapping techniques may address this limitation. Another technique,

myocardial T2* mapping, enables the direct identification and quantification of myocardial

iron in vivo, and has been validated against in vivo histology, making it particularly useful

for the detection of myocardial iron overload as a cause of DCM (109).

Replacement myocardial fibrosis

Replacement myocardial fibrosis occurs in response to cardiac injury (e.g. following

63 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

myocardial infarction), following which apoptosis and myocyte loss occurs, with dead cells

being replaced by collagen-based scar formation. The conventional and gold standard method

to detect myocardial scar is through histological examination(140). However, replacement

myocardial fibrosis can now be detected non-invasively using CMR by exploiting the

alterations in tissue magnetic resonance properties following the administration of an

exogeneous contrast agent.

Gadolinium is paramagnetic, meaning it can become temporarily magnetised when placed in

an external magnetic field, a property conferred by its unpaired electrons. Gadolinium based

contrast agents rapidly diffuse from the vascular space to the extracellular space. In tissues

with uptake of gadolinium, T1 and T2 relaxation times are shortened. Therefore in areas of

replacement fibrosis, with a resulting loss of cells, the extracellular space is enlarged by a

dense, hydrated collagen matrix in which gadolinium accumulates. Upon T1 weighted

imaging these abnormal areas have bright signal.

This imaging is employed through an inversion recovery MRI sequence in which normal

myocardial signal is ‘nulled’ to appear black, leaving the area of fibrosis to appear extremely

bright. Signal detection depends upon contrast between normal and abnormal tissue. The

inversion time is set by the operator to null normal myocardium, so that abnormal

myocardium with focal scar appears bright white. This is known as late gadolinium

enhancement (LGE-CMR). The focal replacement myocardial fibrosis can be detected and

quantified using semi-automated software and validated methods.

CMR detected replacement myocardial fibrosis occurs in approximately one third of patients

with DCM, most frequently in a linear mid-wall distribution(141). CMR detected

64 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

replacement fibrosis has robust histological validation in ischaemic and non-ischaemic

cardiomyopathies(13,142-145). Replacement myocardial fibrosis confers an adverse

prognosis in DCM, predicting mortality including sudden cardiac death irrespective of

LVEF(13,146,147). Therefore a great advantage of CMR over echocardiography in DCM is

that echocardiography does not detect replacement myocardial fibrosis therefore cannot

sufficiently stratify sudden cardiac death risk.

The most commonly used methods to quantify replacement myocardial fibrosis are the full-

width at half maximum (FWHM) and the >2 standard deviation approach (>2SD) (148). The

FWHM method quantifies regions of myocardium with a signal intensity >50% of the

maximally enhanced region while the >2SD approach includes regions with a signal intensity

>2SD above the signal intensity of a reference area of normal myocardium. As yet, there is

no universal agreement on standardized methods to enable comparison of fibrosis

quantification between studies.

The limitations of LGE-CMR are related however to the need to discriminate between normal

and abnormal signal. Diffuse myocardial fibrosis, leading to diffuse expansion of the

extracellular space, a feature of DCM, cannot be detected.

Interstitial fibrosis

Reactive, or interstitial, fibrosis is characterized by the expansion of the cardiac interstitial

space without significant myocyte loss and occurs in response to pressure overload, volume

overload, as well as cardiomyopathies. DCM is associated with major changes in the

extracellular matrix and myocardial collagen deposition in DCM can be diffuse(149). The

gold standard for the detection and quantification of diffuse myocardial fibrosis is through

65 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

myocardial biopsy(140). However, this is limited by sampling error and is associated with

small but significant procedural complication risk (150). Diffuse myocardial fibrosis can now

be detected with CMR using T1 mapping.

Either a Modified Look-Locker Inversion Recovery (MOLLI)(151) sequence or shortened-

MOLLI (ShMOLLI)(152) with balanced steady state free precession (bSSFP) readout are

used. As collagen deposition increases the extracellular space, exogenous contrast

(gadolinium) accumulates, and myocardial tissue longitudinal relaxation time (T1) decreases.

The T1 time is inversely proportional to the concentration of gadolinium in the tissue.

The main biological determinant of an increase in native T1 are oedema and an increase in

the interstitial space (for example due to fibrosis). Native T1 times can be measured pre and

post contrast agents. Native T1 values are influenced by the field strength used (higher native

T1 values at 3 T than at 1.5 T), the specific pulse sequence used, the cardiac phase, and

region of measurement. Normal native T1 values are thus specific to the individual

institution, scanner and protocols and absolute T1 values can only be compared when they

were obtained with the same methods(153). Post contrast T1 times are also influenced by

contrast and patient specific factors such as gadolinium dose, time of measurement post

bolus, patient heart rate, body composition, haematocrit, and clearance rate. Many of these

factors cannot always be standardised leading to potentially unreliable post contrast T1

quantification.

To mitigate this, the extracellular volume fraction (ECV) is calculated. This is derived by

calculating the partition coefficient and correcting for the haematocrit. The partition

coefficient is calculated by measuring the change in T1 before and after equilibration of

66 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

contrast distribution. The ECV provides a quantitative measure of myocardial interstitial

volume. ECV values are therefore more reproducible between different field strengths,

vendors and acquisition techniques than both native and post-contrast T1(153).

CMR T1 and ECV findings correlate with endomyocardial biopsy(154), are prognostically

important(155), and can detect reversible changes in patients with ‘early’ DCM(156).

However, with increasing adoption of these techniques in a variety of disease states, it has

become apparent that T1 mapping is most robust in detecting disease where the pathology is

characterised by high ECV values, such as amyloid disease(157) and to a lesser extent, aortic

stenosis(158). With regards to cardiomyopathies, whilst T1 seems to be able to differentiate

normal and abnormal myocardium, there is considerable overlap between values in control

and disease subjects, and discrimination between cardiomyopathies is poor(159). There has

been one large multicenter study showing the prognostic importance of T1 in DCM, reviewed

further in Section 1.7.2.3.2(160).

Myocardial perfusion

First pass gadolinium imaging (imaging at rest or stress immediately after administration) is

used for myocardial perfusion imaging. Acquisition protocols are usually based on gradient

echo or SSFP sequences combined with parallel imaging to enable ultra-fast acquisition. First

pass stress perfusion CMR can visualize inducible perfusion defects in DCM, often seen as

circumferential areas of subendocardial hypoperfusion. Patients with DCM have been shown

to have reduced myocardial perfusion reserve during stress, but not altered oxygenation(161).

This suggests that the perfusion deficit is not sufficient to cause a reduction in oxygenation

during stress. In addition, as abnormal resting cardiac energetics are not affected by oxygen

administration, the resting impairment in energy metabolism is not secondary to a myocardial

67 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

oxygen deficit, meaning that the impaired perfusion in DCM is likely a reflection of disease

severity and not a pathophysiological driver(161). The degree of myocardial blood flow

impairment in DCM has also been strongly correlated with prognosis (162,163).

1.5.2.1.3 Myocardial strain

Cardiac strain is a unitless measure of myocardial deformation that can occur in 3 planes;

circumferential, radial and longitudinal. Shear strains are the angular strains that occur

between 2 of each of these planes.

Strain measurements are particularly useful in cardiomyopathies as abnormalities may be

detected prior to the development of overt ventricular dysfunction, with reduction in

longitudinal strain occurring prior to impairment in circumferential strain. Analysis of

longitudinal strain using myocardial feature tracking has been shown to predict adverse

outcomes, independently of other predictors such as LVEF and late gadolinium enhancement

(LGE), in patients with DCM (164). Abnormal strain can identify subclinical pathology. In a

study of sarcomeric gene positive/phenotype negative individuals, there were no overt

phenotypic differences compared to healthy controls, only echocardiography detected

abnormalities in myocardial strain discriminated pathology(165).

CMR assessment of strain has historically been challenging but there are novel techniques,

each with their own merits and limitations, such as CMR tissue tagging, feature tracking,

phase-contrast CMR, velocity encoded CMR, DENSE (displacement encoding with

simulated echoes) and SENC (strain encoding)(166). Tagged CMR is the method that has

been most widely studied. In this, selective saturation pre-pulses are used to superimpose a

grid across the field of view across the heart. The grid lines are deformed by myocardial

68 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

contraction, strain, and torsion, allowing direct quantification of myocardial deformation and

strain(167). However it is heavily limited by tag fade in diastole, thereby reducing the amount

of information available in diastole. This is exacerbated at higher field strengths.

DENSE encodes tissue displacement into the phase of the magnetic resonance signal.

However, traditional cine DENSE is limited by fade of the signal throughout the cardiac

cycle. A novel spiral (as opposed to Cartesian) acquisition cine DENSE has been developed

and validated at 1.5T and also at 3T with the introduction of variable flip angles(168-170).

The latter is particularly important as 3T imaging is subject to field inhomogeneities and off

resonance artefacts(168). We have made iterative changes to this cine DENSE sequence to

reduce the breath hold time, making it suitable for patients with DCM, described further in

the methods section of Chapter 4.

1.5.2.1.4 Contractile reserve

Contractile reserve has been assessed as either changes in left ventricular ejection fraction or

changes in global strain measurements following pharmacological or physical stress(171).

The presence of left ventricular contractile reserve in DCM patients, historically assessed

with dobutamine stress echocardiography, is a strong prognostic variable, independent of

resting ventricular indices, even in patients with severely impaired LV function(172-179).

Recently right ventricular contractile reserve has also been shown to be prognostically

important(180). The precision assessment of biventricular ejection fraction with CMR makes

CMR assessed contractile reserve particularly accurate. Reduced myocardial response to

dobutamine indicates reduced adrenergic contractile reserve, thought to relate to the degree of

beta-adrenergic receptor down regulation and desensitization(181). Increasing beta-receptor

down regulation reflects progressive deterioration in left ventricular function(181). Therefore,

69 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

improved contractility during dobutamine administration reflects preserved beta receptor

function, which reflects the potential to show improvement in LV function. The assessment,

prognostic significance, and role of myocardial contractile reserve in DCM is explored

further in Chapter 4.

1.5.3 Limitations of cardiac MRI in dilated cardiomyopathy

Historically, the use of CMR in DCM has been limited by cost and accessibility, though as

the field expands and the number of centres offering CMR increases, these limitations will

subside.

Until recently, patients with intracardiac devices could not undergo CMR, therefore

introducing significant bias into CMR based studies of DCM. Recently devices have been

produced that are MR conditional and patients with these devices can have CMR (182,183),

and in our institution there is increasing experience of safely scanning patients with non-MR

conditional devices.

The acquisition of high quality CMR images requires patient compliance, particularly with

breath holding, and ideally a regular heart rate for ECG gating. A CMR scan with gadolinium

contrast may take up to 60 minutes. Therefore CMR scanning may not be suitable for some

patients with DCM and NYHA class III/IV symptoms or patients with arrhythmia.

Having reviewed the role of CMR in DCM, I now move on to review remodelling and

outcomes in DCM.

70 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.6 Remodelling and Recovery in DCM

1.6.1 Potential for LV remodelling in DCM

There has been longstanding recognition of the potential for improvement in DCM, with

spontaneous improvement in symptoms and potentially complete recovery of LV function

reported in approximately 25-70% of DCM patients (184-195).

However, quantification and predictors of this recovery are poorly defined. Higher estimates

of recovery are reported in more contemporary cohorts, but there is no universal definition of

improvement in LV function, with studies reporting either absolute change in

LVEF(185,187,189,193,194,196,197) (from 5-20%) or improvement above a threshold level,

often set at LVEF 50%(190,193,198), with some studies also reporting reduction in LV

dilation(189,190,193). However, subtle dysfunction in cardiac strain or energetics can

remain, even in the presence of apparently normalized LVEF(198,199).

The terms myocardial recovery and reverse remodelling are often used interchangeably but

they may represent distinct entities(200). In addition, this improvement often occurs in the

presence of medical or device therapy, therefore it is unknown whether LV recovery

represents disease remission(201) or disease cure. In one study of 85 patients with reported

LV recovery, the rate of recurrence of LV dysfunction, even in the presence of ongoing

medical therapy was as high as 38%(202). Baseline age, LV end diastolic diameter, and a

history of diabetes were the only independent predictors of recurrent dysfunction(202). There

are two ongoing randomized studies of withdrawal of medical therapy in recovered DCM

patients that may address the remission versus cure conundrum (Withdrawal of Medication in

Recovered DCM (WrecEF), ClinicalTrials.gov identifier NCT02770443; Therapy withdrawal

71 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

in REcovered Dilated cardiomyopathy trial (TRED), EU Clinical Trials register identifier

2015-005351-27).

1.6.2 Situations in which LV remodelling can occur in DCM

Intuitively, it is plausible that recovery of LV function can occur after withdrawal of an

environmental trigger (e.g. alcohol, virus). It has also recently been shown that recovery is

possible in genetic DCM(191), specifically titin cardiomyopathy, both in response to medical

therapy(85,203) and advanced device therapy such as left ventricular assist device

implantation in apparently “end-stage” patients(204).

There are other specific situations in which LV recovery may occur. Tachycardia induced

cardiomyopathy, developing in response to atrial (atrial fibrillation or tachycardia most

frequently) or ventricular arrhythmia (slow ventricular tachycardia or ventricular ectopy), can

improve upon restoration of sinus rhythm(205). A small case series of 24 patients however

showed that recurrence of tachycardia could lead to a rapid decline in ventricular function,

long after apparent normalization(206), suggesting that LVEF recovery does not equate with

complete normalization of myocardial substrate. In other words, LV reverse remodelling does

not imply molecular myocardial recovery(200). Recovery has also been reported in cases of

acute myocarditis(192,207), peri-partum cardiomyopathy(208), and some forms of

chemotherapy induced cardiomyopathy (trastuzumab, as opposed to anthracycline)(209).

1.6.3 Prognostic implications of LV remodelling

Improved LVEF has also been associated with a survival benefit, though this has mainly been

demonstrated in heart failure cohorts, including ischaemic aetiologies. In the BEST trial

72 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

(beta-Blocker Evaluation of Survival Trial), amongst 2484 patients (not all DCM) with at

least two serial evaluations of LVEF by radionuclide ventriculography, a change of LVEF by

at least 5% was associated with a reduced hazard ratio for all cause mortality (0.62, [0.52-

0.73])(210). Similarly, in the Veterans Affairs Cooperative Vasodilator-Heart Failure Trials

(V-HeFT), serial evaluation of 1446 patients with heart failure showed that a change in LVEF

>5% from baseline to 6 months was the strongest predictor of mortality(186). In the largest

study to date, incorporating data from almost 70,000 heart failure patients across 30 mortality

trials and almost 20,000 heart failure patients across 88 remodelling trials, the odds of

mortality decreased with increasing LVEF and decreasing LV end diastolic and end systolic

volumes(211). In line with the studies before it, a 5% increase in mean LVEF corresponded

to an improvement in survival (OR for all cause mortality with 5% improvement in LVEF

0.86, p=0.013)(211). In a study of over 700 patients with ICD therapy in the MADIT-CRT

trial, LVRR occurred in 25% of patients (defined as >15% reduction in LV end systolic

volume at 1 year follow up) and that this was associated with a reduced risk of heart failure or

death(212). Finally, in the subset of patients with repeat echocardiography imaging in the

Valsartan Heart Failure Trial (Val-HeFT), 321 of 3519 (9.1%) patients with initial LVEF

<35% (including ischaemic aetiologies) improved LVEF to >40% at 1 year follow up, and

recovery was associated with improved survival at 2 year short term follow up (log-rank p

value comparing survival in group with improved LVEF compared to group without

improved LVEF=0.005)(213).

In a small, retrospective DCM specific study of 59 patients, ~37% of 19 patients surviving

beyond 12 years showed LV reverse remodelling and amongst the 33 patients who died or

had a heart transplant, no LV reverse remodelling was noted prior to the event(188). A

cautious conclusion of the study would therefore be that any reverse remodelling, even

73 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

limited, is associated with an improved prognosis. This of course requires replication and

confirmation in a larger prospective study. In a retrospective observational study of 408 DCM

patients, 63 patients with improved LV function (LVEF >50% and normal LV end diastolic

volume) had a greater freedom from death or heart transplant compared to patients without

evidence of recovery (p<0.001)(190).

In summary, recovery has been shown to be possible in the setting of DCM secondary to a

diverse range of aetiologies, albeit to a varying extent. This suggests that the potential for LV

recovery is intrinsically conserved in the setting of cardiac dysfunction (largely irrespective

of aetiology), and even in the presence of apparently severe myocardial dysfunction. The

complicating issue is that recovery is not universal, so identifying which hearts retain this

potential for recovery is a major unmet need.

1.6.4 Predictors of LV remodelling in DCM

Identification of DCM patients with a high probability of recovery has the potential to

improve outcomes, by permitting tailored therapy, stratifying early intensive and advanced

therapy to patients deemed less likely to recover.

There have been limited studies in the contemporary era of medical therapy evaluating

predictors of recovery for all-cause DCM beyond the removal of any initial environmental

trigger.

1.6.4.1 Left ventricular parameters and clinical predictors of LV remodelling

A key study in the field was from the IMAC investigators (Intervention in Myocarditis and

74 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Acute Cardiomyopathy)(192). In total, 373 subjects from 16 centers with recent onset (<6

months) idiopathic DCM or myocarditis with an initial LVEF <40% were followed up for a

mean of 2.2 years, with reassessment of LV function after 6 months. Crucially, 82% of

patients received beta blocker therapy and 91% of patients were on angiotensin converting

enzyme inhibitors (ACEi). Many previous studies in the field predated the widespread use of

these prognostic medications(187). All imaging was performed by echocardiography. Mean

baseline LVEF was 24%, increasing to 40% at 6 months. In multivariable analysis, baseline

left ventricular end diastolic diameter was the strongest predictor of follow up LVEF

(standardized coefficient -0.41, p<0.0001)(192). Other independent predictors of follow up

LVEF were systolic blood pressure (0.18, p=0.001), black race (-0.12, p=0.02) and NYHA

class (-0.11, p=0.04)(192). Interestingly, baseline LVEF did not predict follow up LVEF

(p=0.32)(192). The authors also noted differing recovery profiles stratified by gender and

race, with recovery (LVEF >50%) more likely in white women (38%) and least likely in

black men (15%)(192). Whilst 12% of subjects had an endomyocardial biopsy at baseline, a

limitation of this study was that imaging with CMR was not performed, which may have been

able to identify a reversible inflammatory myocarditis, as well as evaluate the role of

myocardial fibrosis in predicting recovery.

1.6.4.2 Role of CMR detected late gadolinium enhancement in predicting remodelling

A subsequent study of 44 patients with recent onset DCM evaluated CMR predictors of

remodelling, together with serum biomarkers, endomyocardial biopsy, cardiopulmonary

exercise testing and echocardiography(189). In this study, patients had CMR at baseline and

at 1 year follow up, and left ventricular reverse remodelling (LVRR) was defined as an

absolute increase in LVEF of 10% to a final value >35%, together with a reduction in LV end

diastolic dimension at the 1 year mark(189). In total, LVRR was observed in 45% of patients.

75 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Of all variables evaluated including serum biomarkers and myocarditis on biopsy, the

independent predictors of LVRR were the extent of myocardial fibrosis assessed by LGE

CMR (OR 0.67, p=0.008) and higher myocardial oedema ratio (T2 index) assessed on CMR

(OR 1.45, p=0.027)(189). The authors also measured BNP at 3, 6 and 12 months. At 3

months (though not at baseline), plasma BNP was the most powerful predictor of LVRR.

These data suggest that CMR predictors of remodelling were the earliest to identify the

potential for recovery.

Another CMR based study of remodelling in 68 DCM patients (disease onset <2 weeks) also

identified that improvement in LVEF (at 5 months) was inversely correlated with the extent

of LGE(214). Notably, this study excluded patients with suspected myocarditis (abnormal

troponin I or myocardial oedema on CMR), so may be a more accurate study of recovery in

true idiopathic DCM, with estimates of recovery not conflated by inclusion of patients with

myocarditis – a condition with a high degree of reversibility.

Not all studies agree with regards to CMR LGE predicting LVRR. In a Portugese study of

113 DCM patients followed for 7 years, LVRR (defined as absolute LVEF increase of 10%

and decrease in LV diastolic diameter) occurred in approximately one third of patients(194).

On multivariable analysis, only ACEi use was associated with LVRR. CMR-LGE was not a

predictor on univariable analysis. However, only 38 patients had CMR and of these, over

50% had LGE which is higher than conventional estimates, suggesting that their criteria for

identification of mid-wall fibrosis (as opposed to all forms of LGE) was inadequate(194).

Notably, and in line with other studies, baseline LVEF did not predict LVRR.

Similarly, in a separate study of 97 patients with DCM, LVRR (defined as an increase in

76 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

LVEF of 5%) occurred in 71% (n=69) of patients after 1 year follow up(195). On

multivariable analysis, only LV end- diastolic volume and the duration of symptoms on

presentation were independent predictors of 1-year LVEF. Neither baseline LVEF nor the

presence of CMR-LGE (performed in 88 patients) predicted LVRR(195). In another sub-

study of 66 DCM patients who had CMR, LGE was detected both in the patients who did not

experience LVRR as well as in patients with late LVRR (LVEF increase of 10% after 1 year),

although the extent of LGE enhancement was lower in patients who responded compared to

the non-responders(196). This small study, subject to marked ascertainment bias (the original

cohort was >200 patients) suggests that the association between CMR-LGE and LV

remodelling may be more nuanced than previously imagined.

1.6.4.3 Role of left atrial and right ventricular assessment

Many of the studies described limit evaluation of imaging predictors to indices of LV

structure and function only. However, other cardiac structures such as the left atrium(215)

and right may be important predictors of remodelling, as they reflect broader

pathological involvement. A study of 44 patients with DCM without evidence of myocardial

LGE who had CMR assessment of LA volume showed that LA volume was the only

independent predictor of LVRR with a hazard ratio of 0.93 (0.88 to 0.99, p=0.024) (14

patients had LVRR defined as an increase in LVEF to >50% and a net increase of

>20%)(197). This was a small study, the upper confidence interval approaches 1, and the

results whilst interesting, are by no means definitive. Right ventricular function at baseline

has recently been shown to predict LV remodelling at follow up in peri-partum

cardiomyopathy patients(216). This remains unexplored in DCM.

77 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.6.5 Predictors of sustained LV recovery in DCM

The Trieste Heart Muscle Disease Registry is a long running established database of

cardiomyopathy patients at a tertiary referral centre. Study of this cohort has permitted the

evaluation of both predictors of recovery overall(193), as well as predictors of sustained

recovery, implying potential myocardial ‘healing’(190).

In the first study, LVRR (defined as LVEF >50% or absolute LVEF increase by 10%) was

found in 89 of 242 (37%) patients with DCM. Only systolic blood pressure (OR 1.23 per

10mmHg systolic blood pressure, p=0.047) and the absence of left bundle branch block (OR

2.47, p=0.009) predicted LVRR(193). In the second study, published 4 years later with longer

term follow up, in total 15% (n=63) of 408 DCM patients demonstrated ‘apparent healing’,

defined as LVEF >50% and normal indexed LV end diastolic diameter, with 38 of the 63

patients (60%) demonstrating ‘persistent healing’ at long term follow up (mean 103 months).

Amongst those patients who did not demonstrate sustained recovery, LVEF appeared to

deteriorate after approximately 2 years of recovery, but with no clear discriminators of why

this subgroup should have a different clinical course. On univariate analysis, no clinical or

echocardiography imaging parameters predicted LV improvement either at mid-term (approx.

2 years follow up) or long-term follow up (over 8 years follow up)(190). Interestingly, even

amongst the group of patients with persistent apparent healing, 2 of the 38 (5%) patients died

or underwent heart transplant at very long term follow up, despite normalization of LVEF.

Whilst speculative, this suggests that LVRR represents myocardial remission, but not true

healing. In line with this, it has been suggested that in the presence of normalization of

molecular, cellular, myocardial and LV geometric changes permitting the heart to maintain

preserved LV structure or function, the term ‘myocardial recovery’ should only be used to

describe situations associated with freedom from future heart failure events(200) and the term

78 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

‘myocardial remission’ should be used to describe preserved LV structure or function that is

not associated with freedom from adverse events(200). Further support for this notion comes

from a study evaluating outcomes in a cohort of 538 patients with heart failure with repeated

LVEF assessments after primary prevention ICD implantation, in which 40% of patients had

some improvement in LVEF over a mean follow up of 4.9 years (including 25% with LVEF

improvement to >35%) but they remained at risk for appropriate shock therapy(217),

suggesting that the myocardial substrate had not normalised.

In summary, it is clear that there is a distinct subgroup of DCM patients who can undergo

reverse remodelling, either spontaneously or after therapy, and whose clinical course is

associated with reduced adverse events. However, resting LV indices and clinical parameters

have been shown to be inadequate predictors of this LV remodelling and/or recovery. This

may be because resting indices do not reflect the reserve capacity of the myocardium to

remodel. In Chapter 4 therefore, I examine the potential of myocardial contractile reserve, a

dynamic index of LV function, to predict LV remodelling in DCM. In the next section, I

review predictors of prognosis in DCM and highlight how this information could inform risk

stratification in DCM.

79 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.7 Prognosis in DCM and Risk Stratification

1.7.1 Prognosis in DCM

1.7.1.1 Estimates of survival in DCM

The natural history of DCM is difficult to precisely define as the disease may have a variable

latent period, prior to the onset of symptoms. Overall, adult onset DCM remains associated

with an adverse prognosis, though there have been marked improvements in survival with the

advent of medical and device therapy.

Historic estimates of long-term prognosis in DCM showed that the disease was associated

with a 5 year mortality rate of ~65% in the 1960s-70s(218), falling only slightly to 57% in the

1980s(219). In the earlier series, overall mortality approached 80%(218). In both studies, the

majority of deaths occurred within 2 years of diagnosis. The high mortality estimates may be

a reflection of clinical management prior to the introduction of prognostic neuro-hormonal

medication such as ACEi and beta blockers, as well as the referral bias of only severe cases

being referred to the tertiary centres which reported the studies. This theory is supported by a

population based study of outcome in 40 DCM patients in the Olmsted County study

(diagnosed between 1975-1984) in which 5 year mortality was estimated at 20%(220).

Survival estimates from studies in the 1990s showed a marked improvement in outcome, with

a 5 year survival rate of 77% in a study of 201 DCM patients(221). This has been attributed

to a combination of factors, including a tendency for patients in later studies to present

younger and with less advanced disease as well a higher proportion of patients treated with

ACEi and beta blockers(222).

80 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

More recent studies, in the era of device therapy, screening for early disease detection, and

larger prospective cohort studies, show that approximate five year all cause mortality in DCM

is ~20%(12,13,223). Whilst this may reflect a substantial improvement compared to early

case series, DCM remains a condition associated with a poor prognosis. There is a pressing

need to identify improved predictors of outcome to be able to identify the ‘at risk’ patient at

an earlier stage in the natural history of disease, ideally at a stage when intervention can alter

prognosis.

1.7.1.2 The focus on sudden cardiac death

Mortality in DCM consists of both heart failure and sudden cardiac death (SCD), with SCD

accounting for 25-30% of deaths(224,225). Heart failure death occurs as a result of pump

failure. The focus of risk prediction in DCM has been in evaluation of predictors of sudden

cardiac death, in part because there is evidence of benefit of implantable cardiac defibrillator

(ICD) therapy in DCM.

Four randomised controlled trials evaluated ICD use in DCM, independently of cardiac

resynchronisation therapy. The DEFINITE trial randomized 458 patients with DCM (LVEF

<35%, NYHA class 1-3, with non-sustained ventricular tachycardia or ventricular ectopy) to

ICD placement or optimal medical therapy(226). The AMIOVIRT trial had similar

recruitment criteria but randomized 103 patients to ICD or amiodarone(227). In the CAT

(Cardiomyopathy Trial), 104 DCM patients with LVEF <30% and NYHA class 2 or 3 were

randomized to ICD or medical therapy(228). In SCD-HeFT (Sudden Cardiac Death in Heart

Failure Trial), 1210 patients with DCM, LVEF <35% and NYHA class 2 or 3, were

randomized to heart failure therapy plus amiodarone or heart failure therapy plus ICD(229).

81 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

In a meta-analysis of their results, ICD use in DCM was associated with a 26% relative risk

reduction for all-cause mortality (95% CI: 0.59–0.93; p = 0.009)(230). DEFINITE(226) and

SCD-HeFT(229) showed a trend to reduction in all cause mortality with ICD therapy, though

none of the individual trials demonstrated a significant absolute reduction(227,228).

Importantly, separate subanalyses of the SCD-HeFT trial demonstrated a significant

reduction of arrhythmic mortality in DCM (relative risk 0.34; 95% CI: 0.17 to 0.70)(230) and

tachyarrhythmia mortality in NYHA class II patients only (adjusted hazard ratio, 0.40; 95% CI: 0.27 to 0.59)(231), with no reduction in heart failure mortality or death from non-cardiac causes(231). Both the CAT and AMIOVIRT trials were stopped early due to lack of statistical significance in reaching the primary endpoint.

The overall goal therefore is to identify high risk patients who would gain the greatest benefit

from ICD therapy whilst avoiding subjecting low risk patients to a complications from a

procedure with little prognostic benefit. This underlies the importance of risk stratification in

DCM.

In the following sections, I will first review established and novel risk biomarkers in DCM. I

will then review current guidelines for risk stratification in DCM and highlight their

limitations, before concluding with the justification for personalised medicine and multi-

modality risk stratification in DCM.

1.7.2 Established and novel risk stratification biomarkers in adult onset

DCM

DCM can be stratified based on clinical, imaging, biomarker and genetic data. Each domain

contains prognostic information, which I will briefly review, with a focus on novel and

82 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

emerging markers. Paediatric cardiomyopathy has a different risk profile and will not be

reviewed(232).

1.7.2.1 Clinical phenotypes

Symptom status

Symptom status has long been one of the most robust predictors of outcome in DCM.

In a study of 201 DCM patients followed up for ~4.5 years(221), severity and duration of

symptoms, as well as indices of LV function, predicted outcome. Specifically, symptoms of

pulmonary or peripheral oedema and syncope were associated with a worse outcome. Five

year survival ranged from 90% in patients with no symptoms and better cardiac function, to

53% in patients with severe symptoms for a long duration and cardiac impairment. In a

separate study of 232 patients with cardiomyopathy referred for cardiac transplantation

evaluation, 40% with DCM, NYHA class was one of only 3 independent predictors of

survival (p<0.0001), the others being pulmonary capillary wedge pressure and plasma atrial

natriuretic factor(233). Syncope has also been shown to predict sudden cardiac death in

patients with ischaemic and non-ischaemic cardiomyopathy(234).

The importance of symptom status has since been borne out in an independent study,

showing that survival is better in asymptomatic compared to symptomatic DCM

patients(235). In this study, 220 asymptomatic patients within a cohort of 747 DCM patients

in the Heart Muscle Disease Registry of Trieste were followed up for ~ 9 years. Patients in a

higher NYHA class had a worse prognosis for the end points of sudden death, heart transplant

and cardiovascular hospitalization compared to asymptomatic patients (p < 0.001 among all

groups). There was no significant difference on sub-analysis of consistently asymptomatic

patients compared to those with a previous history of heart failure symptoms but now

83 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

asymptomatic on medical therapy. This suggests that improved symptom status may be a

dynamic risk marker, such that the risk of adverse events over time decreases as symptoms

improve. Finally, NYHA class was also an independent predictor of all cause mortality, heart

failure death, heart failure hospitalization, and cardiac transplantation, but not sudden cardiac

death, in a large contemporary outcome study of 472 patients with DCM(13).

Age and gender

In some studies, age has been associated with outcome, but this has not been consistent

across historic case series(219,236). In the Olmstead county population study, older age was

associated with reduced survival with an adjusted hazard ratio for mortality of 1.59 for each

10 years increase in age(220). In the most recent study of outcomes in over 472 patients with

DCM, a one year increase in age was independently associated with all cause mortality with a

hazard ratio of 1.03 (p=0.001)(13).

Gender based differences in outcome have been noted in cardiovascular disease, particularly

ischemic heart disease, but somewhat surprisingly, gender has not been an important

independent predictor of survival in DCM after LVEF and NYHA class are accounted

for(13,113).

Co-morbidities

Systolic blood pressure has been associated with outcome in DCM to a varying

extent(237,238). It has not always been evaluated in historic studies of outcome in

DCM(233). Most recently a more favourable prognosis was observed in patients with a

higher blood pressure (hazard ratio for all cause mortality per 1mmHg increase in systolic

84 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

blood pressure 0.98, p=0.01)(13). Conversely, in a study of 87 patients with DCM,

individuals with a lower mean arterial pressure were less likely to survive, though this was

not an independent predictor of survival(238). Other important co-morbidities in DCM

include diabetes(239) and renal dysfunction. Co-morbidities also present a competing risk of

death in patients at risk of SCD. For example, patients with an ICD can die due to co-morbid

conditions, which may become more advanced with increasing age(240).

Other

Historically, haemodynamic abnormalities such as pulmonary capillary wedge pressure,

cardiac index, and central venous pressure have been associated with poor prognosis, but

their utility in long term outcome prediction is limited and will not be reviewed further(224).

1.7.2.2 ECG data

Ventricular arrhythmias, particularly ventricular ectopy and non-sustained ventricular

tachycardia (NSVT), are common in DCM, with NSVT identified in up to 50% of patients in

a prospective series of 74 DCM cases(241). In that study, patients with DCM and LVEF

<40% who had frequent episodes of NSVT were at a higher risk of SCD(241). Whether this

is a marker of an impaired left ventricle or an independent prognostic indicator remains to be

definitively evaluated(242), though there evidence to suggest that NSVT is of particular

prognostic importance amongst patients with mildly dilated cardiomyopathy in predicting

death or cardiac transplant (NSVT HR 2.21, p=0.047), where LVEF did not predict

outcome(243).

A major issue in evaluating this area is ascertainment bias, whereby even if a patient

85 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

undergoes ambulatory monitoring for 24-48 hours as is common in clinical studies, it is

plausible that arrhythmia can go undetected. Longer periods of monitoring are carried out in

sicker patients, in whom ventricular arrhythmias are more likely to exist, therefore

reinforcing the notion that arrhythmia portends a worse outcome. Another complicating issue

is the variable definition of arrhythmia, for example a study of 218 DCM patients in whom

some form of ventricular arrhythmia (including ectopy) was found in 94% of patients, but

ventricular tachycardia found in only 27% of patients(244). By this definition, increasing

severity of ventricular arrhythmia was associated with reduced survival (p<0.001), but the

threshold of clinically important NVST was not clear. A study of a mixed cohort of ischaemic

and non-ischaemic cardiomyopathy patients suggested that ventricular tachycardia on

ambulatory monitoring was independently associated with total mortality (p=0.008) and

sudden death (p=0.003), with a tachycardia frequency of >0.088 events/hour being associated

with a 3 fold increase in mortality rate(245). However, the inclusion of patients with

ischaemic heart disease complicates the applicability of these findings to DCM patients.

Even recent studies evaluating the prognostic importance of ventricular arrhythmia do not

adequately address this issue. In a study of 285 DCM patients, an arrhythmic phenotype was

an independent predictor of SCD, sustained ventricular tachycardia or ventricular

fibrillation(246). However, the definition of arrhythmia was broad and included any one of

unexplained syncope, rapid NSVT (at least 5 beats over 150bpm), >1000 premature

ventricular contractions/24 hours, or >50 ventricular couplets/24 hours(246).

In summary, ventricular tachycardia is likely to be of prognostic importance in DCM but

there is insufficient evidence as yet to state which patients with ventricular arrhythmia are at

greatest risk of sudden death.

86 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Other ECG data may be of prognostic value in DCM, such as microvolt T wave alternans

(TWA), but despite evidence that it may have incremental value in predicting SCD risk in

patients with LVEF<35%(247-249), it is not a parameter widely used in clinical practice.

Abnormal T wave alternans provides an assessment of the temporal and spatial heterogeneity

of repolarization, which could trigger ventricular arrhythmias. In a meta-analysis of 45

studies of SCD risk stratification in DCM including 6088 patients, the highest odds ratio to

predict SCD was for T wave alternans (4.66, 95% CI 2.55 to 8.53) and fragmented QRS (OR

6.73, 95% CI 3.85 to 11.76)(250). The predictive value of the TWA test is higher when

performed on patients taking beta blockers. Other variables evaluated including parameters of

autonomic instability were not predictive of SCD. The incremental prognostic value of TWA

over and above LVEF has not been evaluated in a randomized controlled trial.

1.7.2.3 Imaging phenotypes

1.7.2.3.1 Established imaging markers

LVEF

Left ventricular ejection fraction (LVEF) is a robust predictor of outcomes in DCM, with

LVEF <35% associated with the highest risk of sudden cardiac death, as reflected in major

national and international guidelines recommending ICD therapy(251-253). Much of the

background data to support a cut off of LVEF <35% originates from studies in the setting of

ischaemic heart disease(229,254,255), but LVEF <35% is also of importance in SCD risk

stratification in DCM as shown in the SCD-HeFT trial as previously described(229).

In the two major early DCM mortality outcome studies of 104 patients followed up at the

87 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Mayo Clinic(218) and 169 patients followed up at the Hammersmith Hospital London(219),

the only independent predictor for outcome was the severity of LV impairment at

presentation, as assessed by increased cardiothoracic ratio on chest radiography and LV

dilation and LVEF respectively. In the Olmstead county population study, lower LVEF was

independently associated with reduced 1 year survival, with an adjusted hazard ratio of 1.90

per 10% decrease in LVEF(220). A number of other studies have reinforced the importance

of LVEF in predicting outcome in DCM(245,256). In more contemporary studies, LVEF

<30% remains a strong independent predictor of death, heart transplantation, and arrhythmia

in asymptomatic patients(235). In all-comer DCM patients, LVEF independently predicts all

cause mortality (HR 0.97 per 1% increase in LVEF, p=0.007), as well as composites of

cardiovascular mortality or cardiac transplant (HR per 1% increase in LVEF 0.96, p<0.001),

sudden cardiac death or ventricular arrhythmia (HR per 1% increase in LVEF 0.97, p=0.005),

and heart failure death, hospitalisations or heart transplant (HR per 1% increase in LVEF

0.95, p<0.001)(13).

The limitations of LVEF however are that it is a load dependent measure and can be affected

by heart rate variability as well as concomitant pathology e.g. mitral regurgitation. There is

also variability in LVEF measurement depending on imaging modality chosen. Furthermore,

as outlined in the review of remodelling in DCM, LVEF is known to vary with disease

course. The degree to which a single static measurement, taken at a variable timepoint in the

natural history of disease, can capture all future risk is unclear.

RVEF

Right ventricular dysfunction has also been shown to be an independent predictor of outcome

in DCM, with the presence of right ventricular dilation (assessed by echocardiography) being

88 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

associated with reduced long term survival in a study of 67 patients with DCM(257). In a

larger prospective study of 250 patients with DCM who underwent CMR, the presence of

right ventricular systolic dysfunction (RVEF <45%, present in 34% of patients) was an

independent predictor of all cause mortality/cardiac transplantation (hazard ratio 4.20,

p<0.001), cardiovascular mortality, cardiac transplantation and heart failure death/heart

failure hospitalization(113). Interestingly given the limitations of LVEF, in a cohort of 314

patients (164 ischaemic and 150 non-ischaemic cardiomyopathy), RV dysfunction (RVEF

<45%) was an independent predictor of adverse events (sudden cardiac death and ICD

therapy) in patients with LVEF >35% (adjusted HR4.2, p=0.02)(258).

Left atrial size

Left atrial (LA) size has also been linked to prognosis in DCM. In a study of 144 patients

with DCM followed up for the composite end point of cardiovascular death or cardiac

transplantation, only left atrial dimension index as measured by echocardiography and

pulmonary wedge pressure were independent predictors of outcome(259). LVEF was not

associated with outcome after adjusting for LA dimension, thereby being one of the earliest

studies to highlight the role of the LA as a independent imaging biomarker in DCM(259). In

a CMR based study of 483 DCM patients, followed up for the primary endpoint of all-cause

mortality or cardiac transplantation, indexed LA volume was an independent predictor of the

endpoint (HR per 10mL/m2 increase in indexed LA volume 1.08, p=0.022)(127). LA volume

was also independently associated with the secondary endpoints of cardiovascular mortality

or cardiac transplant and heart failure death/heart failure hospitalizations/cardiac

transplant(127).

89 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.7.2.3.2 Novel imaging markers

Replacement myocardial fibrosis

Myocardial fibrosis acts a substrate for ventricular arrhythmias, an important cause of sudden

cardiac death(260). There is now an increasing body of evidence demonstrating the

importance of mid-wall fibrosis, detected on late gadolinium enhancement CMR (LGE-

CMR), as a prognostic indicator in DCM.

Assomull et al reported the first study to demonstrate that LGE-CMR midwall fibrosis had

prognostic importance in DCM(141). In this study of 101 patients with DCM, mid wall

fibrosis (present in 35% of patients), was associated with a hazard ratio of 3.4 (p=0.01) for

the risk of all cause mortality and cardiovascular hospitalisation. On prespecified secondary

endpoint analysis, midwall fibrosis also predicted sudden cardiac death or ventricular

tachycardia, with a hazard ratio of 5.2 (p=0.03)(141).

Subsequently, Lehrke et al evaluated the significance of LGE-CMR in 184 patients with

DCM(261). The mean follow up time was relatively short at 1.9 years, and the composite end

point was cardiac death, heart failure hospitalisation, or appropriate ICD discharge. Overall,

39% of patients had LGE-CMR and both the presence and extent of fibrosis, predicted the

composite end point (adjusted hazard ratio for presence of LGE-CMR 3.4, 95% CI 1.26 -

9)(261).

Focusing specifically on arrhythmia, Marra et al evaluated whether LGE-CMR predicts

sustained ventricular tachycardia, appropriate ICD activation, ventricular fibrillation and

sudden cardiac death in 137 patients with DCM, followed up for a median 3 years for the

primary composite end point (262). In this study, the presence, but not the extent of LGE-

90 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

CMR predicted malignant arrhythmia, with an adjusted hazard ratio of 3.8 (95% CI 1.3-10.4;

p = 0.01)(262).

In the largest single centre study to date, Gulati et al, evaluated the impact of LGE-CMR mid

wall fibrosis on all cause mortality and an arrhythmic composite endpoint in a prospective

study of 472 patients with DCM(13). Patients were followed for a median of 5.3 years. LGE-

CMR midwall fibrosis was independently associated with sudden cardiac death, appropriate

ICD shock and non fatal ventricular tachycardia or ventricular fibrillation (hazard ratio 4.61

[95% CI, 2.75-7.74], p < 0.001)(13). LGE-CMR midwall fibrosis was also independently associated with cardiovascular mortality or cardiac transplantation (hazard ratio, 3.22 [95%

CI, 1.95-5.31], p <0 .001)(13).

Aside from these clinical studies, there are a number of smaller studies evaluating LGE-CMR

in DCM, but they have variable definition of LGE-CMR fibrosis, variable end point

definition and variable follow up time(147). In this context, further evidence for the role of

LGE-CMR in predicting sudden cardiac death risk came from a meta-analysis of eleven

studies comprising 1105 patients with ischaemic and non-ischaemic cardiomyopathy. In a

median follow up of 41 months, 207 patients had ventricular arrhythmias, which was

associated with LGE-CMR fibrosis (relative risk 4.33, [95% confidence interval (CI) 2.98 -

6.29])(263). In the subgroup of 3 studies with DCM patients only (n=277), the relative risk

was 3.79 (95% confidence interval 1.20 -11.94)(263).

A recent DCM specific systematic review and meta-analysis of data from 29 studies,

encompassing 2948 patients with DCM, evaluated the association between LGE-myocardial

fibrosis and the risk for ventricular tachyarrhythmia in patients with DCM(264). Patients

were followed for an average of 3 years, and the composite arrhythmic endpoint consisted of

91 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

sustained ventricular tachycardia, appropriate ICD activation or sudden cardiac death. Mid-

wall fibrosis was present in 44% of patients, and the arrhythmic composite was met by 21%

of patients in the LGE positive group, compared to 4.7% of patients in the LGE negative

group(264). Overall, the combined odds ratio for the arrhythmic composite endpoint was 4.3

(95% confidence interval 3.3 to 5.8), demonstrating that LGE was a strong independent

predictor of arrhythmic events in patients with DCM.

In summary, the evidence to date regarding the prognostic importance of mid-wall fibrosis in

DCM is compelling, with LGE-CMR an independent predictor of heart failure

hospitalisation, malignant arrhythmias, all cause mortality and sudden cardiac death.

However, there remain limitations and unanswered questions. For example, it is not yet

established what threshold of fibrosis confers arrhythmic risk. Most analyses to date have

evaluated the presence or absence of fibrosis as a binary variable. Quantitative reporting of

CMR detected fibrosis is variable and there is no standardised methodology(265). We also

know that peripheral biomarkers of fibrosis exist, but we do not yet know what the

significance of LGE-CMR is over and above peripheral fibrosis markers. With a few notable

exceptions such as CMR_GUIDE (ClinicalTrials.gov Identifier: NCT01918215), there are

limited randomised trials in this field to evaluate whether clinical management informed by

knowledge of CMR-LGE changes outcomes in DCM. Furthermore, the presence and extent

of mid-wall fibrosis varies between patients with the same degree of ventricular impairment.

The genetic susceptibility to the development and maintenance of a fibrotic substrate has also

not been fully elucidated.

Interstitial myocardial fibrosis

Areas of intermediate, or interstitial fibrosis, can be detected through CMR T1 mapping and

92 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

the calculation of extracellular volume, as previously outlined. Its prognostic value in DCM

has been evaluated in a limited number of studies.

In a cohort of 139 patients with cardiomyopathy (59 with non-ischaemic aetiology), CMR T1

mapping was performed prior to ICD implantation(266). After follow up for a mean of 1.2

years, 23 patients (18%) met the primary end point of ICD therapy or sustained ventricular

arrhythmia. In multivariable analysis, native T1 was associated with the primary endpoint

(HR for every 10ms increase in T1 1.10, 95% CI 1.04 to 1.16). T1 mapping was a stronger

predictor of outcome than quantification of replacement fibrosis in the same patients(266).

Clearly this cohort was small, the follow up was relatively short and inclusion of ICD therapy

in the endpoint may over-estimate the SCD risk associated with interstitial fibrosis.

In the largest study of T1 mapping in DCM to date, 713 patients with DCM underwent T1

mapping (native, post contrast and ECV measurement) and were followed up for a median of

22 months for the primary endpoint of all cause mortality. In total, 28 patients (4.4%) met the

primary endpoint. On univariate analysis, all T1 indices were associated with outcome. On

multivariate analysis, adjusting for RVEF and LGE, native T1 was independently associated

with mortality, with an adjusted hazard ratio of 1.1 per 10ms change in T1(160). Whilst

potentially important for the field, highlighting the importance of interstitial fibrosis over and

above replacement fibrosis, this study was limited by the short follow up and low event rate,

meaning the authors could not evaluate all conventional predictors of outcome in DCM

alongside evaluation of T1 indices. In post-hoc analysis, the authors report that the

combination of LVEF <35% or LGE with native T1 did not improve predictive value, but

this analysis may have been insufficiently powered. Therefore interstitial fibrosis remains an

area of interest for future, likely multi-centre studies of outcome in DCM.

93 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Having reviewed clinical and imaging biomarkers in DCM, I now briefly review the potential

prognostic role of serum biomarkers in DCM.

1.7.2.4 Evaluation of blood biomarkers

1.7.2.4.1 Established blood biomarkers

Natriuretic peptides

A number of structurally similar natriuretic peptides have been identified: atrial natriuretic

peptide (ANP), urodilantin (an isoform of ANP), B-type natriuretic peptide (BNP), C-type

natriuretic peptide and Dendroaspis natriuretic peptide. Of these, B-type natriuretic peptide

and the biologically inert precursor cleavage product N-terminal proBNP (NTproBNP),

largely secreted by cardiomyocytes, regulate salt and water handling and cardiac

remodelling(267). Elevated levels reflect ventricular strain, predominantly stretch. They are

the most well studied biomarkers in heart failure and provide excellent biomarkers for heart

failure diagnosis, prognosis, and treatment, independently of left ventricular ejection

fraction(268-271).

Across the heart failure spectrum, from acute to chronic, and asymptomatic to symptomatic,

BNP and NTproBNP have been consistently shown to predict prognosis. A systematic review

of 19 studies evaluating the ability of BNP to predict death or cardiovascular events found

that each 100pg/mL increase in BNP was associated with a 35% increase in the relative risk

of death in heart failure patients (p=0.096)(270).

In the ADHERE study of 48,629 acutely decompensated heart failure patients, of whom

19,544 patients had reduced ejection fraction, there was a linear relationship between

94 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

increasing admission BNP and in hospital mortality, which persisted after adjustment for

conventional risk factors(272).

In a direct comparison between NTproBNP and BNP in 3916 patients with chronic heart

failure in the Valsartan Heart Failure Trial, both were found to be independent predictors of

all cause mortality (area under ROC curve BNP 0.665, NTproBNP 0.679), mortality and

morbidity, and hospitalization for heart failure, with NTproBNP being marginally superior

for the latter two endpoints(273).

There has been limited evaluation of the role of natriuretic peptides in DCM specifically. The

use of NT-proBNP in guiding heart failure therapy in a randomised trial of 151 patients with

chronic left ventricular systolic dysfunction (LVEF <30%) showed that NT-proBNP guided

care was associated with a reduction in cardiovascular events, improved quality of life and a

favourable cardiac structural remodelling profile(274). In this cohort, 33% of patients in the

NT-proBNP and 24% of patients in the standard care arm had non ischaemic

cardiomyopathy, suggesting that NT-proBNP may have prognostic value in DCM as well as

all cause heart failure.

There are of course limitations in the use of these peptides. BNP has high levels of

physiological variability(275) and can also be affected by comorbid conditions of obesity,

sepsis and renal failure, the latter a particularly common comorbidity in heart failure patients.

Cardiac troponins

The cardiac troponin proteins are located in myocytes and regulate contraction. Cardiac

troponin, formed of three subunits (troponin-C, -I and –T), is released in response to myocyte

95 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

necrosis and is the established biomarker for detection of acute myocardial infarction.

Patients with heart failure may have chronically elevated levels, which may be related to

elevated ventricular preload leading to increased myocardial strain(276). The degree of

troponin elevation indicates severity of myocardial damage, and may be higher in patients

with ischaemic as opposed to non-ischaemic cardiomyopathies(277), but does not

conclusively point to underlying aetiology. However, in the heart failure population, elevated

troponin is a poor prognostic indicator(278). In a large study of over 80,000 patients

hospitalised for acute decompensated heart failure, 6.2% of patients had an elevated troponin.

The adjusted odds ratio for death in patients with an elevated troponin was 2.55 (95%

confidence interval, 2.24 to 2.89; p<0.001)(278).

Within the studies limited to patients with DCM (as opposed to heart failure, all causes),

elevated troponin I in 310 patients with DCM predicted all cause mortality, together with

ECG QRS duration, NYHA class and systolic blood pressure(279). In another smaller study

of 95 patients with DCM, with a median follow up of 4.1 years, the composite endpoint of

cardiovascular mortality, heart transplantation, or ICD implantation was met by 37% of

patients with an elevated troponin compared to 8% of patients with a non-elevated troponin

(p<0.01)(280).

1.7.2.4.2 Novel blood biomarkers

Many other biomarkers have been associated with prognosis in heart failure and to a lesser

extent, in DCM. The end stage phenotype of dilated cardiomyopathy reflects insults across a

diverse range of pathways and cellular processes. Perturbations in many of these pathways

may precede overt phenotypic development. The breadth of novel biomarkers reflects a range

96 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

of pathogenic processes that may occur in the development of DCM, covering inflammation,

oxidative stress, extracellular matrix turnover, neurohormonal activation, myocyte injury and

myocyte stress (Table 1-6).

Not all of these biomarkers have been linked to outcome in DCM and they will not be

reviewed further as they are not evaluated within this thesis. To conclude this section on risk

stratification biomarkers in DCM, I will next review genetic biomarkers of risk DCM.

97 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Table 1-6: Biomarkers linked to heart failure classified by biological function. Not all biomarkers listed

have been proven to predict outcome.

Myocyte injury Other

Cardiac-specific troponins Neutrophil gelatinase associated lipocalin (NGAL) I and T Cystatin C Creatine kinase MB fraction Growth differentiation factor 15 Myocyte stress Transforming growth factor-beta1 Brain natriuretic peptide Osteopontin N-terminal pro-brain natriuretic peptide Osteoglycin ST2/ interleukin-33 Syndecan-1 Inflammation Syndecan-4 C-reactive protein Extracellular-matrix remodelling/fibrosis Tumor necrosis factor Matrix metalloproteinases 1, 2, 3, 8, 9 Fas (APO-1) Tissue inhibitors of metalloproteinases 1, 4 Interleukins 1, 6,10, 18 Collagen propeptides: Oxidative stress • C-terminal propeptide of procollagen Myeloperoxidase type I Oxidized low-density lipoproteins • N-terminal propeptide of procollagen type I Neurohormones • N-terminal propeptide of procollagen Norepinephrine type III Renin • C-terminal telopeptide of collagen type I Angiotensin II Galectin-3 Aldosterone Osteoprotegerin Arginine vasopressin

Endothelin

98 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.7.2.5 Evaluation of genetic determinants of risk in DCM

1.7.2.5.1 Evidence of genetic contribution to risk in DCM

As outlined, approximately 20-30% of DCM has a familial origin and a genetic basis has

been identified in up to 40% of these cases(5). In studies to date, DCM appears

morphologically similar irrespective of the underlying genetic cause. However, there is some

evidence that genetic variants modulate arrhythmic risk in DCM.

In a large study of predictors of arrhythmic risk in 285 DCM patients, it was found that a

family history of sudden cardiac death, sustained VT or ventricular fibrillation, was an

independent predictor of the arrhythmic endpoint (SCD, VT, VF) on long term follow up

(mean ~9 years)(246). This suggests that there may be a genetic contribution to the risk of

sudden cardiac death in DCM.

The 2013 report by the American Heart Association and leading American cardiac societies

(ACCF/HRS/AHA/ASE/HFSA/SCAI/SCCT/SCMR) on criteria for ICD implantation goes so

far as to recommend ICD implantation in patients with LVEF >35% (who therefore would

not meet conventional criteria) if they have familial DCM which has been associated with

SCD, regardless of the specific genetic mutation(112).

Regardless of a family history of SCD, it remains to be established whether having genetic or

familial DCM confers an adverse prognosis, irrespective of genetic variant. The evidence,

though limited, suggests that genetic DCM per se is not associated with poor outcomes

compared to non genetic DCM. In a study of 213 DCM patients, no differences in the rate of

adverse events was detected between the 70 (33%) patients with genetic or familial DCM and

99 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

patients without such a history(95). A pathogenic mutation was identified in only 16 cases so

genotype-phenotype study was not performed.

Similarly, in a study of paediatric DCM patients (647 idiopathic DCM, 223 familial DCM),

there was no difference in outcomes (all cause mortality or cardiac transplant) on

multivariable analysis between children with familial or non familial DCM (281). In fact, on

univariable analysis, children with familial DCM had a trend to lower risk of death compared

to children with idiopathic DCM (HR 0.64, p=0.06). This of course may be confounded by

earlier disease detection and earlier institution of prognostic medication in familial cases and

may not reflect genetic contribution to risk at all.

Large prospective studies documenting the correlation between gene mutations and SCD risk

are lacking(282). Therefore to discriminate further, and for genetic risk stratification to be

useful, evaluation of well studied genotype-arrhythmic phenotype correlations is required. I

will focus the next section on rare genetic variants only.

1.7.2.5.2 Potential rare genetic variant biomarkers of arrhythmic risk

A meta-analysis of 48 genotype-phenotype studies including 8097 DCM patients with

mutations in LMNA, PLN, RBM20, MYBPC3, MYH7, TNNT2 and TNNI3 found that the

frequency of ventricular arrhythmia was highest for patients with variants in LMNA (50%)

and PLN (43%)(283).

DCM patients with variants in the LMNA gene are at high risk for malignant ventricular

arrhythmias(284). After studying 269 LMNA carriers for a median follow up time of 43

months, this risk was shown to be highest for patients with a history of NSVT, LVEF <45%,

100 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

male gender and truncating variants (49). For patients with DCM, conduction disease and

identified LMNA variants, clinical guidance suggests that an implantable cardiac defibrillator

should be considered in preference to a conventional pacemaker due to the identified

genotype-phenotype correlation of an increased risk of malignant arrhythmias and sudden

cardiac death (49).

Therefore genetic risk stratification has a role in LMNA DCM but there are unanswered

questions. It is unclear whether all patients with DCM should be tested for LMNA variants, or

only those with a family history of disease. The degree to which the risk associated with

LMNA variants is modulated by other known SCD risk factors in DCM, for example CMR-

LGE enhancement, has not been studied. Finally, whether the presence of a LMNA variant

should be factored into decisions on ICD therapy for primary prevention in DCM also needs

to be evaluated in a large prospective study.

Variants in the PLN gene, common in the Netherlands due to the presence of the founder

mutation, are also notable for their association with an arrhythmic phenotype(63). In

comprehensive analysis of a family with the PLN-R14Del, it was found that all mutation

carriers who may or may not have developed overt ventricular impairment had evidence of

attenuated R wave amplitudes on ECG and that these areas corresponded to the regions of

LGE on CMR(285). As with previous studies, there was an increased association with sudden

cardiac death(285). In a large prospective Dutch cohort study of 418 patients with DCM, 43

patients had variants in PLN and 19 patients had variants in LMNA which were both

associated with a malignant ventricular arrhythmias and end stage heart failure (hazard ratio

for death, heart transplant, malignant arrhythmia in patient with a mutation 2.0, 95% CI 1.4-

3.0)(66). Therefore there is a strong evidence of the link between the PLN founder mutation

101 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

and arrhythmic risk, but the low frequency of these variants outside of the founder population

may limit their utility in genetic risk stratification.

One of the early phenotype studies of TTNtv cardiomyopathy (n=42) suggested TTNtv DCM

was associated with sustained ventricular tachycardia at baseline (OR 6.7, p=0.001) and that

this association was robust to adjustment for LVEF(84). This has not yet been replicated in

other studies and the long term association with arrhythmic risk remains to be established

before TTNtv can be used as a genetic biomarker of arrhythmic risk in DCM.

As well as TTNtv, variants in other sarcomeric genes (MYH6, MYH7, MYBPC3 and TNNT2)

have, in combination, been shown to be associated with an increased risk of cardiovascular

death/heart transplant/VF in DCM in subjects after the age of 50 years(286). However, this

analysis was based on the presence of 24 variants in total therefore there were insufficient

events (3 sudden deaths) to adequately discriminate which genes were driving the findings

and whether they could discriminate SCD from heart failure death. This interesting

observation requires further study in a large prospective cohort of patients with DCM.

Recently, truncating variants in the gene encoding filamin C (FLNCtv) have been linked to

an arrhythmic cardiomyopathy phenotype(78). On evaluation of 2,877 patients with inherited

cardiac conditions, 23 truncating mutations in FLNC were identified. These patients had

previously been diagnosed with DCM or arrhythmic or restrictive cardiomyopathies. A

further 54 mutation carriers were identified amongst 121 screened relatives. Notably, the

presence of FLNCtv was associated with ventricular arrhythmias (found in 82% of cases) and

sudden cardiac death (40 cases in 21 out of 28 families)(78). Whilst these findings have not

yet been replicated in a distinct clinical cohort, the results are compelling and provide

102 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

evidence to support screening for FLNCtv in patients with DCM to guide risk stratification.

Other variants in desmosomal proteins, classically associated with ARVC, have also been

identified in DCM which may confer an increased risk of arrhythmic adverse events(287).

However, because there is considerable phenotypic overlap between DCM and ARVC and

the recognition of a high level of background variation at a population level in some of these

genes, the importance of desmosomal gene mutations in predicting sudden cardiac death risk

in DCM requires further clarification. Similarly, variants in SCN5A (sodium channel) have

previously been linked to DCM and associated with arrhythmia(288). However, variation in

SCN5A is now known to be relatively common and the role of SCN5A as a DCM gene has

been questioned(14,67). Variants in SCN5A may still have a role in the modulation of

arrhythmic risk in patients with DCM and this requires further investigation.

In summary, there are a number of potential genetic biomarkers of arrhythmic risk in DCM,

but the most robust association with relevance to a global cohort of patients with DCM has

been defined for patients with variants in the LMNA gene. The unmet need is to evaluate the

additive effect of genetic risk prediction in the context of established and emerging non-

genetic prognostic markers in DCM.

Having reviewed potential clinical, imaging, biochemical and genetic prognostic biomarkers

in DCM, I conclude this Introduction with a review of our current approach to risk

stratification in DCM. I review the limitations of our current guidelines and show where the

work in this PhD seeks to address some of these limitations.

103 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.7.3 Current risk stratification in DCM and limitations

1.7.3.1 Current guidelines for ICD implantation

Current guidelines for primary prevention ICD implantation from major international

societies are outlined in Table 1-7.

Table 1-7: Current guidelines for primary prevention ICD implantation from European and American

guidelines. OMT= optimal medical therapy; NSVT= non sustained ventricular tachycardia; LVEF= left

ventricular ejection fraction; NYHA = New York Heart Association functional class; SCD=sudden

cardiac death.

Society/Overall criteria DCM specific criteria Class of Level of Recommendation evidence ESC(289) Non ischaemic aetiology I B

DCM with LMNA IIa B • LVEF ≤35% mutation and risk factors • NYHA classes II or III (LVEF <45% at • At least 3 months of presentation, NSVT on OMT ambulatory ECG, non- • Expected survival missense mutation) for ≥1 year with good functional status

AHA/ACC/HRS(252) Non ischaemic aetiology I B

• LVEF ≤35% Familial cardiomyopathy IIb C • NYHA classes II or III with a history of SCD

In both European and American guidelines, LVEF is the major determinant of whether or not

to implant an ICD.

104 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.7.3.2 Limitations of current guidelines in risk stratification of patients with DCM

LVEF below 35% is currently the major variable driving selection of patients with DCM for

implantable cardiac defibrillator therapy. The limitations of this approach have been

demonstrated in a recent landmark randomised control trial, the DANISH study(12).

In this study, 1,116 patients with symptomatic DCM (LVEF <35%) were randomized to

either an ICD or standard clinical care. Patients in either arm could have cardiac

resynchronization therapy (CRT); in the final analysis ~60% of patients in both groups had

CRT. The patients were followed for a median of 67.6 months and the primary result showed

that ICD therapy was not associated with a significant reduction in all cause mortality

compared to usual clinical care(12). It is worth noting that the ICD group showed reduction

in incidence of sudden cardiac death by half (4.3% vs 8.2%, HR 0.50, p=0.005), though this

was not the primary endpoint. In other words, although SCD was reduced, mortality risk was

converted from SCD to pump failure deaths. This leads us to one of three possible

conclusions.

One conclusion is that ICD therapy is not beneficial in DCM. There is insufficient evidence

to support this. In a recent meta-analysis of 6 randomised controlled trials enrolling 2970

patients with DCM, evaluating the use of ICDs for primary prevention, ICDs were associated

with a reduction in all-cause mortality (pooled HR 0.77, 95% CI 0.64-0.91)(290). After

excluding trials where patients had CRT therapy, ICDs were still associated with a reduction

in all cause mortality (pooled HR 0.76, 95% CI 0.62-0.94).

The second possible conclusion is that the benefit of ICD therapy in the modern era of

optimal medical therapy (including CRT) is reduced compared to the original trials. This has

105 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

been supported by a recent meta-analysis of 5 trials including 2992 patients comparing ICD

therapy to medical therapy for primary prevention in DCM. Compared to medical therapy,

ICD therapy was associated with a reduction in overall mortality (pooled OR 0.77, p=0.006)

(291). Secondary endpoints of sudden cardiac death were also reduced in the ICD arms but

other cardiovascular and non-cardiovascular causes of death did not differ between groups.

The investigators noted that the DCM cohort in DANISH had the highest compliance with

optimal medical therapy of all trials included and it was the only study with a high rate of

CRT implantation, which may account for the discrepant data(291). In the previously referred

to meta-analysis of 6 RCTs, a direct comparison of the COMPANION and DANISH-CRT

subgroup trials, comparing ICD plus CRT plus medical therapy to CRT plus medical therapy,

there was a non-significant benefit of ICD therapy (pooled HR 0.70, 95% CI 0.39-1.26)(290).

The third possibility is that ICDs are beneficial in DCM, but current guidelines are not

adequately identifying patients most likely to benefit. There is a growing body of evidence to

support this. Even in subgroup analysis of the cohort of patients in the DANISH trial, there

was an interaction of survival benefit with ICD use in younger patients with DCM, indicative

of benefit of ICD in specific DCM populations.

The rate of appropriate ICD activation in those patients who do meet current criteria is under

20%, again suggesting that current ICD guidelines do not target those most at risk of

SCD(229,292). Observational data prior to the DANISH trial had already shown that adverse

cardiac events occur in patients who do not meet criteria for ICD implantation(293-296). In

the Oregon Sudden Unexpected Death Study, all cases of sudden cardiac death were

evaluated for left ventricular ejection fraction. Amongst a population of 660,486, there were

714 cases of sudden cardiac death. Of these, LV function was assessed in 121 cases and

106 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

found to be normal in 58 cases, representing 48% of sudden cardiac death cases. LV function

was mild to moderately impaired in 27 cases, representing 22% of sudden cardiac death

cases. In this study, individuals with sudden cardiac death and normal cardiac function were

more likely to be female, younger, had a higher prevalence of seizure disorder and a lower

prevalence of coronary artery disease. Whilst this study is now over 10 years old, it

demonstrated the identification of sudden cardiac death risk factors beyond LVEF.

Whilst the absolute number of sudden cardiac death events in individuals with normal or

mildly impaired cardiac function is small(297), the number of patients in the subgroup of

mild-moderately impaired cardiac function is large compared to the number of patients with

severely impaired cardiac function(293). Therefore current risk stratification, based on

LVEF, does not manage the risk of death in a potentially large cohort of patients with DCM.

Subsequent studies have evaluated non LVEF predictors of sudden cardiac death, such as

QRS duration, T wave alternans(298), signal averaged ECG, and heart rate variability.

However, whilst many of these techniques showed promise in the early trials, they have not

been replicated or have limited negative predictive value, particularly in the setting of

patients with reduced LVEF(299). Few have been evaluated in the context of DCM as much

of the early work in this field has been in ischaemic cardiomyopathy.

Goldberger et al performed a meta-analysis evaluating these predictors in DCM specifically,

including data on 6,088 patients from 45 studies(250). Baroreflex sensitivity, heart rate

turbulence, heart rate variability, left ventricular end-diastolic dimension, LVEF,

electrophysiology study, NSVT, left bundle branch block, signal-averaged electrocardiogram,

fragmented QRS, QRS-T angle, and T-wave alternans were included. Each of the predictors

107 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

had only modest predictive value for sudden cardiac death in DCM patients. After adjustment

for missing data and publication bias, LVEF had the highest odds ratio for prediction of

sudden cardiac death (LVEF OR: 2.73, 95% CI: 1.99 to 3.76, p < 0.001), but even then the

odds ratio was modest for clinical risk prediction to be based on LVEF alone(250). This

limits the utility of each predictor in isolation, and in such a situation, multimarker

approaches to risk stratification are warranted.

It has also been shown that sudden cardiac death risk is clinical context specific. For

example, in a retrospective multicentre study evaluating the outcomes of heart failure patients

with LVEF <45% who met and did not meet National Institute of Clinical Excellence (NICE)

criteria for ICD implantation, patients with diabetes who did not meet the criteria had a

similar absolute risk of sudden cardiac death or appropriate ICD shock as patients without

diabetes who met the criteria(300).

Together, these data therefore suggest that current ICD guidelines are not appropriately

targeting the population most likely to benefit from therapy. Given the limitations of

established guidelines, there is a need to integrate the multi-dimensional data we now have

with regards to the specific risk in DCM, particularly for patients who fall outside the

guideline criteria. As the absolute incidence of sudden cardiac death is still relatively low in

patients with mildly impaired cardiac function, any risk stratification tool will need to be

powerful to be able to identify patients at risk. It will also need to be discriminatory enough

to avoid creating an undue burden on the health care system.

This will involve evaluation of multiple phenotypes with the ability to discriminate risk

amongst homogenous groups, for example, discriminating risk between patients in the same

108 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

LVEF strata. One potentially discriminatory phenotype is mid-wall fibrosis on cardiac MRI.

A clinical trial addressing the utility of LGE mid-wall fibrosis data in patients who do not

currently qualify for an ICD is currently underway (CMR_GUIDE, ClinicalTrials.gov

Identifier: NCT01918215). This is designed to evaluate whether, amongst heart failure

patients with mild-moderate left ventricular systolic dysfunction (LVEF 36-50%), the use of

a CMR guided management strategy is superior to a conservative strategy of standard care. In

this, patients with mild-moderate left ventricular systolic dysfunction and evidence of LGE

midwall fibrosis on CMR will be randomised to either ICD or implantable loop recorder

insertion and followed up for 3 years for the primary endpoint of sudden cardiac death or

ventricular tachycardia leading to syncope. The trial is due to report in 2020.

In summary, current risk stratification guidelines in DCM rely heavily on LVEF. LVEF is an

excellent predictor of sudden cardiac death in a subgroup of patients at risk of sudden cardiac

death, but in isolation, LVEF has low sensitivity and specificity for the prognostic

stratification of sudden cardiac death risk in DCM. Impaired LVEF reflects established

disease, at which point risk modulation may be limited. Reliance on only one or limited

phenotypic markers ignores the biological complexity of DCM. Our understanding of the

pathobiology of DCM now incorporates advanced imaging, genetic and biomarker data. As

outlined, we have an understanding of risk based on each of these domains. Critically, many

of these novel biomarkers, in particular genetic changes, can be identified early in the natural

history of DCM. Early identification of the ‘at risk’ patient maximises the effectiveness of

risk prevention, with the downstream potential to modify the disease course. Use of multiple

risk markers in combination could provide a more robust prediction of events(301).

109 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

1.7.4 Personalised medicine in DCM

Personalised medicine ‘implies a tailored approach to patients that offers more effective

therapy for each individual, reduces risks and avoids unnecessary treatments or diagnostic

interventions’ (ESC Position Paper on Personalised Cardiovascular Medicine)(302).

The potential application of personalised medicine in DCM is vast, particularly with the

increasing understanding of the genetic basis of disease, as well as prognostically important

endophenotypes detected through advanced imaging. Yet this brings challenges, as by design,

the DCM populations in clinical studies cannot be entirely representative of all possible

subgroups of disease (either by genetic aetiology, clinical demographics or stage in disease

course for example). Although clinical, biological and imaging features are helpful in

determining risk in large populations, there is no seamless transition to the assessment of

prognosis for an individual patient. It is in this junction that personalized medicine fits,

offering the ability to stratify a phenotypically homogenous disease into distinct subgroups,

through clinical, biochemical, imaging, and genomic markers. The utility of these subgroups

would depend on their clinical robustness and the demonstration of benefit of earlier

detection of an at risk phenotype or introduction of tailored therapy leading to improved

outcomes.

At present there is no integrated patient stratification in DCM that incorporates clinical,

genetic and non-invasive imaging markers for the diagnosis, risk stratification, and

management of patients with DCM. Through an integrated understanding of the genetic basis

and phenotypic manifestation of DCM, the work in this thesis seeks to address this need.

110 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

AIMS, OBJECTIVES AND HYPOTHESES

Research aim:

Evaluate whether integrated assessment of genotype and phenotype data in DCM can

improve understanding of DCM pathogenesis, inform patient stratification, and identify

predictors of remodelling and outcomes.

Objectives:

1. To define the genetic architecture of dilated cardiomyopathy in the context of

background population genetic variation (chapter 2).

2. To describe the clinical manifestations of titin cardiomyopathy and evaluate

phenotypic modifiers, with the purpose of informing patient stratification and

identifying novel insights into disease pathogenesis (chapter 3).

3. Evaluate imaging predictors of remodelling in dilated cardiomyopathy, including

myocardial contractile reserve assessed by low-dose dobutamine stress CMR,

myocardial strain, interstitial fibrosis and replacement myocardial fibrosis (chapter 4).

4. To define the clinical outcomes associated with titin cardiomyopathy to inform patient

stratification by genotype in DCM (chapter 5).

Hypotheses addressed in Chapter 2

- Truncating variants in the titin gene are the commonest genetic contributor to DCM.

- Evaluation of the genetic architecture of DCM, comparing the burden of rare variants in a

DCM cohort to population variation, will identify genes associated with DCM and will

better define putative gene-DCM associations.

111 Template by Friedman & Morgan 2014 Word Chapter 1: Introduction: Dilated Cardiomyopathy

Hypotheses addressed in Chapter 3

- The cardiovascular phenotype of dilated cardiomyopathy in patients with titin truncating

variants is distinct to that observed in patients without titin truncating variants.

- Alcohol consumption is an environmental modifier of left ventricular ejection fraction in

patients with DCM and truncating variants in titin.

Hypotheses addressed in Chapter 4

- Baseline left ventricular contractile reserve in patients with recent onset DCM is

predictive of LV remodelling at 1 year.

- Baseline circumferential and longitudinal myocardial strain, replacement myocardial

fibrosis, interstitial myocardial fibrosis, and RV contractile reserve in patients with recent

onset DCM are each predictive of LV remodelling at 1 year.

- An adequate baseline left ventricular contractile reserve in patients with DCM reflects a

lower level of interstitial myocardial fibrosis when compared to patients with poor

baseline left ventricular contractile reserve.

Hypotheses addressed in Chapter 5

- Patients with DCM and titin truncating variants have an adverse event profile for the

composite endpoint of cardiovascular mortality, major heart failure and major arrhythmic

events, compared to patients with DCM without titin truncating variants.

- The effect on outcome of titin truncating variants in patients with DCM is modified in the

presence of additional environmental variables, including male gender, mid-wall

myocardial fibrosis, arrhythmia, and moderate excess alcohol consumption.

112 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2 EVA LU ATIN G TH E GENETIC

ARCHITECTURE OF DILATED

CARDIOMYOPATHY

2.1 Aims and hypotheses

The overall aim in this chapter is to evaluate the burden of rare genetic variation in suggested

DCM genes in a large cohort of DCM patients compared to reference population variation, in

order to evaluate the genetic architecture of DCM. The primary purpose of this chapter is for

the results to inform the phenotype analysis in Chapter 3. The secondary aim is to define

DCM disease gene associations that may in the future be informative for patient stratification.

Specific hypotheses are as follows:

- Truncating variants in the titin gene are the commonest genetic contributor to DCM.

- Evaluation of the genetic architecture of DCM, comparing the burden of rare variants in a

DCM cohort to population variation, will identify genes associated with DCM and will

better define putative gene-DCM associations.

2.2 Background

The genetic architecture of dilated cardiomyopathy (DCM) is complex, characterised by a

113 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

large number of low frequency variants with marked locus and allelic heterogeneity.

The advent of next generation sequencing technologies has made DNA sequencing more time

and cost efficient. This has led to a rapid increase in the number of reported disease-gene

associations. To date, almost 70 genes have been implicated as causative for DCM, but many

of these associations have been made through studies with small case numbers, in the absence

of segregation or functional data, and without robust population specific controls.

The large scale exome sequencing project ExAC (Exome Aggregation Consortium) has

demonstrated a relatively high level of rare genetic variation at a population level(15). This

discovery provides an opportunity to evaluate putative DCM disease-gene associations

against population level background genetic variation. The ExAC dataset has recently been

used to refine rare variant thresholds, predominantly in cancer (BRCA genes), and to a lesser

degree respiratory disease (primary ciliary dyskinesia) and hypertrophic

cardiomyopathy(303). The authors found that of the previously published pathogenic variants

studied, most had an allele frequency less than 0.01% in ExAC, and the majority much lower

than that. Pathogenic variants with allele frequencies greater than 0.01% were often founder

mutations or reflected mutation hotspots. Similarly, a statistical framework for calculating

disease specific allele frequency (accounting for disease prevalence, genetic and allelic

heterogeneity, inheritance, penetrance and sampling variance in reference datasets), has

shown that allele frequency cut offs well below 0.1% are justified for a variety of human

diseases including for example hypertrophic cardiomyopathy, in which the maximum

credible population allele frequency for any causative variant is 4.0x10-5(24).

Over the last 20 years (1996-2015), there have been 68 DCM genes reported in Human Gene

114 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Mutation Database (HGMD), 44 of those in the last 10 years alone(34). Understanding which

associations are robust and have true clinical significance is a key challenge faced by

clinicians and geneticists. By leveraging the power of the ExAC dataset to inform us about

background population level rare genetic variation, our group has recently demonstrated the

potential this approach could have in understanding the genetic architecture of

cardiomyopathies, particularly hypertrophic cardiomyopathy(14,304). Whilst this study also

evaluated genetic variation in DCM genes in up to 1315 cases with DCM, this analysis was

limited to major DCM genes as many recently reported putative DCM genes were not

represented on older gene panels, for example ZBTB17(60) and BAG3(52). Furthermore, the

analysis was also limited by lower case numbers for the minor DCM genes or those more

recently reported, a limitation I set out to address in this study with sequencing of DCM

patients using a cardiac gene panel which included up to 57 of the 68 genes reported to be

associated with DCM(305).

2.3 Methods

2.3.1 Study population

2.3.1.1 DCM cohort

The primary DCM study population for genetic analysis was prospectively enrolled in the

National Institute for Health Research (NIHR) Royal Brompton Hospital Cardiovascular

Biobank project. This cohort consisted of consecutive referrals from a dedicated

cardiomyopathy service at the Royal Brompton Hospital, London, consecutive referrals to the

CMR unit, both from within the Royal Brompton Hospital and a network of thirty regional

hospitals, as well as patients with DCM who underwent echocardiography due to

contraindications to CMR. Patients were referred for diagnostic evaluation, screening and

115 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

assessment of severity of DCM. All patients were prospectively enrolled for research

purposes.

All patients underwent cardiac phenotyping with either cardiovascular magnetic resonance

scanning (CMR) or transthoracic echocardiography. For patients with CMR, DCM was

diagnosed based on evidence of left ventricular dilation and systolic impairment with

reference to age, gender, and body surface area adjusted nomograms(116). For patients with

echocardiography, DCM was diagnosed in the presence of left ventricular end-diastolic

diameter (LVEDd) > 117% of that predicted for age and body surface and left ventricular

ejection fraction (LVEF) < 45% and/or fractional shortening (FS) < 25%(9).

Exclusion criteria for DCM included a history of uncontrolled systemic hypertension,

coronary artery disease (>50% stenosis in one or more major epicardial arteries or previous

percutaneous coronary intervention or coronary artery bypass grafting), chronic excess

alcohol consumption meeting criteria for alcoholic cardiomyopathy (>80g/day for more than

5 years(97)), systemic disease known to cause DCM, pericardial disease, congenital heart

disease, infiltrative disorders (e.g. sarcoidosis) or significant primary valvular

disease(8,306,307).

2.3.1.2 Healthy volunteer cohort

A cohort of healthy volunteers (HVOL) was recruited at the Hammersmith Hospital, London,

via advertisement for the UK Digital Heart Project at Imperial College London. These

individuals had no history of medical illness, were not taking regular medication, and did not

have evidence of cardiac structural or functional impairment on CMR scanning.

116 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

All participants gave written informed consent and the study was approved by the relevant

regional research ethics committees.

2.3.2 ExAC population reference dataset

The ExAC dataset is a collation of multiple exome sequencing datasets, at the time of

analysis numbering 60,706 unrelated individuals. The raw data from the different common

disease and population genetics studies have been re-processed through the same pipeline and

jointly variant called. Whilst not a control dataset, the cohort is not expected to be

specifically enriched for individuals with DCM. The ExAC dataset was downloaded from the

ExAC website (http://exac.broadinstitute.org) [version 0.3, Jan 2015]. Only ExAC variants

with a PASS filter were used in this study.

2.3.3 Coverage analysis between cases or controls and ExAC

In order to control for technical differences between platforms that might introduce bias, the

coverage of each gene in ExAC was adjusted for when making the frequency comparison.

We adjust for the number of variants per individuals sequenced, and not simply the number

of variants as a function of all 60,706 individuals.

2.3.4 Gene selection

The Human Gene Mutation Database(34) (HGMD, professional version 2015.3) was

interrogated for all genes with variants associated with DCM reported between 1996-2015.

This generated a list of 68 genes (Table 2-1).

117 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Table 2-1: Genes with variants reported to be associated with DCM in HGMD between 1996-2015

Year Genes Year Genes 1996 DMD 2008 MYPN 1998 ACTC1 CHRM2 1999 DES DSG2 LMNA LAMA2 2000 CTF1 2009 ANKRD1 DSP RBM20 MYH7 NEXN SGCD 2010 TAZ TNNT2 NEBL 2001 TPM1 SYNE1 2002 TTN TCF21 CSRP3 2011 GATAD1 MYBPC3 MURC TCAP DSC2 VCL DOLK 2003 ACTN2 JUP LDB3 PKP2 PLN BAG3 2004 ABCC9 TXNRD2 SCN5A 2013 ISL1 TNNC1 EMD TNNI3 LAMP2 2005 MYH6 PRDM16 TMPO NPPA 2006 CRYAB SGCB FKTN GATA4 DNAJC19 FHOD3 FLT1 ACTA1 PSEN2 2014 GATA6 SYNM CASQ2 PSEN1 RAF1 2007 FOXD4 2015 GATA5 PDLIM3 NKX2-5 ILK ZBTB17 LAMA4

Of these, 57 genes with the strongest evidence of association with DCM are included on the

TruSight Cardio panel (Table 2-2 shows gene and transcript). As RBM20 has poor coverage

in ExAC (mean percentage of sample bases covered at x10=14.5%, mean percentage of

sample bases covered at x20=13.3%), it was excluded from the DCM versus ExAC

comparison, and instead the burden of variants in RBM20 in the DCM cohort was compared

to the control cohort sequenced with the same assay and platform.

118 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Table 2-2: Genes and transcript linked to DCM included on Tru Sight Cardio sequencing panel.

Gene Transcript Gene Transcript Gene Transcript

ABCC9 ENST00000261201 ILK ENST00000299421 PRDM16 ENST00000270722

ACTA1 ENST00000366684 JUP ENST00000393930 RAF1 ENST00000251849

ACTC1 ENST00000290378 LAMA2 ENST00000421865 RBM20 ENST00000369519

ACTN2 ENST00000366578 LAMA4 ENST00000424408 SCN5A ENST00000333535

ANKRD1 ENST00000371697 LAMP2 ENST00000200639 SGCB ENST00000381431

BAG3 ENST00000369085 LDB3 ENST00000361373 SGCD ENST00000337851

CASQ2 ENST00000261448 LMNA ENST00000368300 TAZ ENST00000299328

CRYAB ENST00000526180 MURC ENST00000307584 TBX20 ENST00000408931

CSRP3 ENST00000533783 MYBPC3 ENST00000545968 TBX5 ENST00000310346

DES ENST00000373960 MYH6 ENST00000356287 TCAP ENST00000309889

DMD ENST00000357033 MYH7 ENST00000355349 TMPO ENST00000266732

DNAJC19 ENST00000382564 MYL2 ENST00000228841 TNNC1 ENST00000232975

DOLK ENST00000372586 MYPN ENST00000358913 TNNI3 ENST00000344887

DSC2 ENST00000280904 NEXN ENST00000334785 TNNT2 ENST00000367318

DSG2 ENST00000261590 NKX2-5 ENST00000329198 TPM1 ENST00000403994

DSP ENST00000379802 NPPA ENST00000376480 TTN ENST00000589042

EMD ENST00000369842 PDLIM3 ENST00000284770 TXNRD2 ENST00000400521

FKTN ENST00000223528 PKP2 ENST00000070846 VCL ENST00000211998

GATAD1 ENST00000287957 PLN ENST00000357525 ZBTB17 ENST00000375743

2.3.5 Targeted sequencing

All DCM patients and HVOLs underwent next generation targeted sequencing of 174 genes

associated with inherited cardiac conditions using the custom developed Illumina TruSight

Cardio Sequencing kit for sequence capture and the Illumina MiSeq and NextSeq platforms

for sequencing. The genes sequenced are listed in Appendix Tables 1 and 2.

The TruSight Cardio gene panel was developed by our group to provide comprehensive, high

119 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

coverage sequencing across a range of genes associated with inherited cardiac conditions.

The development, optimization and performance of this panel have been outlined

previously(305). The targeted panel has high levels of coverage of target regions (>99.8% at

≥ 20× read depth on MiSeq platform, >99.9% at ≥ 20× read depth on NextSeq platform)

compared to whole exome sequencing (WES), deep WES, or whole genome sequencing

(88.1, 99.2 and 99.3% coverage at ≥ 20× read depth respectively across the target regions).

Laboratory scientists performed the sequencing. A brief outline of the steps involved follows.

Firstly, targeted DNA libraries were prepared according to manufacturers’ protocols. A DNA

sequencing library consists of short fragments of DNA with appropriate sequencing platform

specific adaptors ligated. To create libraries, genomic DNA was extracted and purified using

standard automated approaches and quality and quantity were assessed by agarose gel

electrophoresis and/or fluorometry (Qubit, Life Technologies). DNA was then fragmented,

using an enzymatic fragmentation process whereby a transposase enzyme fragments the DNA

and inserts the adapter sequence into dsDNA (a process called tagmentation).

For targeted resequencing, custom hybridisation capture probes were designed to enrich the

sequencing libraries made from whole genome samples to target the selected genes

implicated in cardiac conditions. DNA capture probes were used for Illumina assays. Target

capture was performed according to the manufacturer’s protocols using a biotin/streptavidin

pull down method. Following elution from the magnetic beads, the targeted DNA sequencing

library was then amplified using primers that contained a unique 8 barcode

sequence to enable sample multiplexing. The target DNA was then amplified. For the

Illumina platforms, solid phase amplification was used whereby the denatured DNA library

was annealed to adaptors on a slide to undergo bridge amplification to form clusters of

120 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

amplified target DNA for sequencing. Next generation sequencing then occurred using

sequencing by synthesis.

Sequencing by synthesis for the Illumina platform occurs through cyclic reversible

termination. The primed DNA template is bound to the flow cell and a DNA polymerase adds

just one modified nucleotide (which is fluorescently labelled with a terminator). DNA

synthesis terminates after the addition of this one nucleotide. Fluorescence imaging is then

performed to identify the incorporated nucleotide. Then the terminating group and fluorescent

dye are cleaved and the process is repeated. This process occurs in parallel across the

numerous clusters so the recorded fluorescence is a consensus across all the identical

templates in one cluster at that time. This generates approximately 100-150bp reads.

2.3.6 General and platform specific bioinformatic analysis

Variants were annotated and filtered according to customised in house bioinformatics

pipelines as previously described(305).

A team of bioinfomaticians run the bioinformatics pipeline. In brief, the general steps

involved are outlined. The output of NGS reads are in a Fastq format. These undergo per

sample quality control, prior to mapping against a reference human genome. The subsequent

SAM and BAM files undergo pre-processing, in which duplicates are marked, quality scores

are recalibrated, read groups are added or replaced (which read belongs to which sample),

reads are sorted by start position and local realignment around indels occurs (because false

positive single nucleotide polymorphisms (SNPs) can occur near indels). Variant calling is

done with GATK Haplotype Caller or Unified Genotyper. The variant call format (VCF) file

is then uploaded to our in house database and variant filtering and annotation occurs.

121 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Variant calling can detect either the true SNP or can reflect errors resulting from the library

preparation, base calling error in sequencing, mapping error and misalignment (particularly

around indels and repetitive regions), or even an error in the reference genome. Variant

filtering is used to remove sequencing errors and keep the true variants. The filters can be

applied through a software specific calling algorithm, and include assessment of coverage,

allelic balance, strand bias, quality scores, mapping qualities, and read position (for example

proximity to an indel).

Specifically, as applied to the output of the MiSeq or NextSeq platforms: Reads were

demultiplexed (allowing zero mismatches) with Illumina's software and quality was checked

in FastQC v.0.10.1108. Low quality (<20) reads/bases were trimmed using PrinSeq

v0.20.4109 and the good quality reads were aligned to hg19 reference using BWA

v0.7.10110. Marking duplicates reads, local realignment around indels and base quality score

recalibration processes were done in Picard v1.115105 and GATK v3.2-2104. Alignment

summary metrics, callability and coverage reports were calculated using Picard, Samtools

v0.1.18106, Bedtools v2.11.2107 and in house perl scripts. A subset file was created

(ontarget), based on reads mapping quality > 8 and we used this “ontarget” file to make

consistent variant calls in GATK HaplotypeCaller and UnifiedGenotyper. Bases covered by

at least 10 reads with a mapping quality ≥10 and base quality ≥20 were denoted as “callable”,

i.e. adequately covered for variant calling with recommended GATK parameters.

2.3.7 Variant filtering

Poor quality variants that did not meet sequencing quality filters were excluded: any variant

with a quality of depth score less than 4, a read depth less than 10 and an allelic balance less

than 0.20. Following these filtering steps, a list of variants present was generated. To account

122 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

for the differing bioinformatics pipelines used between our cohort and ExAC, any variants

found to be in excess in our cohorts but not present at all in ExAC would be further curated to

check whether these variants had been excluded in the ExAC quality filtering processes

(VQSR sensitivity filters and hard filters). Any such variants would be removed from the

analysis in the DCM and control cohorts to avoid false positive enrichment. The overall

variant filtering process including sequencing control steps is outlined in Figure 2-1.

Selection of Variant filtering Variant filtering Analysis target genes Step 1 Step 2

HGMD review Sequencing filters Burden tesVng All DCM genes Manual curaVon of Quality of depth >4 1996-2015 common variants versus Allelic balance >20% ExAC FAIL VSQR and Coverage >10x hard filters 68 genes Literature review Novel or rare (MAF <0.0001) protein altering Final variant list variants only 57 genes included in our customised sequencing Variant list generated panel (total 174 genes)

Figure 2-1: Overview of methods for variant curation for burden testing in DCM study cohort and

healthy volunteers

2.3.8 Burden testing

The frequency of rare variants (defined as a minor allele frequency (MAF) <0.0001 in the

ExAC dataset) was calculated for each gene in the DCM, HVOL and ExAC cohorts. This

MAF cut off is a disease specific frequency threshold, informed by disease prevalence,

penetrance, and allelic contribution to disease(24), outlined further in Section 3.3.3.2. Only

protein altering variants (missense, nonsense, frameshift, essential splice site and inframe

indels) were assessed, with separate calculations for truncating (nonsense, frameshift and

essential splice site) and non-truncating variants (missense and inframe indels).

123 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Burden testing was performed comparing the burden of rare variants in each gene between

Caucasian DCM patients and Non-Finnish Europeans (NFE) in ExAC, and between

Caucasian HVOLs and NFE in ExAC using Fisher’s exact test (one sided for cases vs ExAC,

two sided for HVOLs vs ExAC), with a significance level of p<0.05, adjusted with

Bonferroni correction for multiple testing (57 genes, p<0.0009).

2.3.9 Sample size calculation

To establish if the DCM cohort was adequately powered for testing the burden of disease

causing variants in DCM genes, the sample size calculation for case and control cohorts was

performed based on the frequency of truncating variants in TTN, the main gene of interest.

The power.prop.test() in R was used, with an estimated frequency of TTNtv 15% in

cases and 1% in controls(84), with an alpha level of 0.05 and power of 0.8, with one sided

testing. This resulted in a sample size of at least 54 in each group. For other genes, where the

excess burden in cases compared to controls was not known, I performed a simulated power

calculation based on an estimated 1% frequency of variation in controls, showing the change

in power for a given DCM gene burden, at differing sample sizes (Figure 2-2).

124 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Figure 2-2: Simulated power calculation for burden testing. X axis shows the estimated burden of the

DCM disease gene. A background population burden of 1% was assumed. The y axis shows the change in

power (for a significance level of 0.05) with differing sample sizes (colours as per legend). For a sample

size of 500 (black curve), there is 80% power to detect a DCM gene burden of 3.7% (dashed line),

compared to 1% burden in controls/population (2.7% excess in cases).

2.4 Results

2.4.1 Cohort overview

The final cohorts for burden testing consisted of 647 Caucasian DCM patients, compared to

33,370 Non-Finnish European participants in the ExAC dataset and 588 Caucasian healthy

volunteers (Figure 2-3).

Of the primary DCM study cohort, there were 443 men (68%), mean age at recruitment was

54.5 years (range 14-88 years) and the majority of the cohort were NYHA class 1 or 2

125 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

(n=514, 79%). The majority of the cohort (n=587, 91%) underwent CMR at recruitment and

mean left ventricular ejection fraction was 39% (standard deviation, 12.3%).

Of the healthy volunteers, there were 260 men (44%), mean age at recruitment was 39.7 years

(range 18-76 years) and the mean left ventricular ejection fraction was 66% (s.d. 5.2%).

Figure 2-3: Overview of cohorts in burden testing

126 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.4.2 Coverage comparison between DCM cases and ExAC

The mean percentage of sample bases covered at 30x in the DCM cohort and 10x in ExAC

per selected gene is shown in Figure 2-4. Four genes included in the burden analysis have

<70% of sample bases covered at 10x in ExAC: TPM1, TAZ, NKX2-5, and DNAJC19. All

except two genes (GATAD1 and LDB3) in the DCM cohort have >99% of sample bases

covered at ≥ 30x. Mean coverage at 10x and 20x in ExAC and mean and median coverage at

10x and 30x in the DCM cohort is shown in Table 2-3. Generally, coverage is better in the

selected genes in the DCM cohort compared to the ExAC samples.

Figure 2-4: Coverage plot showing the mean percentage of sample bases covered at 10x in the DCM

cohort (black dots) and 10x in the ExAC cohort (red dots) for the 57 selected genes. RBM20 is shown for

comparison, though it has not been included in the burden analysis.

127 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Table 2-3: Coverage comparison between DCM and ExAC datasets. Mean percentage of sample bases

per gene covered at 10x or 30x in the DCM cohort and at 10x or 20x in ExAC.

Gene DCM Mean DCM DCM Mean DCM ExAC mean ExAC mean 10x Median 10x 30x Median 30x 10x 20x ABCC9 100.0 100.0 100.0 100.0 95.7 92.0 ACTA1 100.0 100.0 99.8 100.0 97.4 91.3 ACTC1 100.0 100.0 99.9 100.0 99.2 96.4 ACTN2 100.0 100.0 99.9 100.0 91.8 86.8 ANKRD1 100.0 100.0 100.0 100.0 98.2 92.9 BAG3 100.0 100.0 99.9 100.0 92.5 87.2 CASQ2 100.0 100.0 100.0 100.0 98.1 90.0 CRYAB 100.0 100.0 100.0 100.0 75.6 63.7 CSRP3 100.0 100.0 100.0 100.0 99.8 97.7 DES 100.0 100.0 99.6 100.0 80.6 69.3 DMD 100.0 100.0 100.0 100.0 90.2 76.3 DNAJC19 100.0 100.0 100.0 100.0 48.8 41.5 DOLK 100.0 100.0 100.0 100.0 99.6 98.5 DSC2 100.0 100.0 100.0 100.0 92.4 87.1 DSG2 100.0 100.0 100.0 100.0 94.1 88.9 DSP 100.0 100.0 100.0 100.0 98.4 95.0 EMD 100.0 100.0 99.7 100.0 86.3 79.4 FKTN 100.0 100.0 100.0 100.0 78.1 74.1 GATAD1 99.9 100.0 97.8 100.0 74.6 72.3 ILK 100.0 100.0 100.0 100.0 94.5 90.9 JUP 100.0 100.0 99.5 100.0 94.0 84.4 LAMA2 100.0 100.0 100.0 100.0 97.3 92.6 LAMA4 100.0 100.0 100.0 100.0 95.6 90.9 LAMP2 100.0 100.0 99.9 100.0 94.9 87.5 LDB3 98.7 99.0 96.4 96.7 89.4 82.3 LMNA 100.0 100.0 99.7 100.0 74.3 57.8 MURC 100.0 100.0 100.0 100.0 98.7 95.7 MYBPC3 100.0 100.0 99.4 100.0 73.4 50.8 MYH6 100.0 100.0 99.6 100.0 98.9 95.2 MYH7 99.9 100.0 99.6 100.0 99.3 96.3 MYL2 100.0 100.0 99.9 100.0 99.1 95.0

128 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Gene DCM Mean DCM DCM Mean DCM ExAC mean ExAC mean 10x Median 10x 30x Median 30x 10x 20x MYPN 100.0 100.0 100.0 100.0 72.5 69.4 NEXN 100.0 100.0 100.0 100.0 90.0 81.4 NKX2-5 100.0 100.0 100.0 100.0 66.7 46.5 NPPA 100.0 100.0 100.0 100.0 98.8 97.1 PDLIM3 100.0 100.0 100.0 100.0 86.1 78.5 PKP2 100.0 100.0 99.9 100.0 94.6 89.2 PLN 100.0 100.0 100.0 100.0 99.7 97.2 PRDM16 100.0 100.0 99.3 100.0 82.0 74.4 RAF1 100.0 100.0 100.0 100.0 95.6 92.9 RBM20 100.0 100.0 99.9 100.0 14.6 13.4 SCN5A 100.0 100.0 99.7 100.0 92.2 78.4 SGCB 100.0 100.0 99.8 100.0 82.9 79.3 SGCD 100.0 100.0 100.0 100.0 72.2 57.0 TAZ 100.0 100.0 99.1 100.0 64.9 52.3 TBX20 100.0 100.0 100.0 100.0 93.1 80.6 TBX5 100.0 100.0 100.0 100.0 93.5 85.7 TCAP 100.0 100.0 99.6 100.0 95.7 79.1 TMPO 100.0 100.0 100.0 100.0 95.0 89.5 TNNC1 100.0 100.0 99.4 100.0 92.4 84.1 TNNI3 100.0 100.0 99.8 100.0 81.0 65.5 TNNT2 100.0 100.0 99.9 100.0 84.9 75.5 TPM1 100.0 100.0 100.0 100.0 65.4 57.9 TTN 100.0 100.0 99.9 100.0 91.4 85.4 TXNRD2 100.0 100.0 99.3 100.0 81.0 73.3 VCL 100.0 100.0 100.0 100.0 94.4 87.6 ZBTB17 100.0 100.0 99.9 100.0 79.6 67.7

129 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.4.3 Excluded variants

In total, variants in 61 participants were excluded (Table 2-4). Seven patients and four

HVOLs carried a non-truncating variant in the gene MURC

(c.702_722delAAGGCAGTCAGGAGAGAGGCT) that had been excluded in the ExAC

quality filtering processes. The remaining variants were excluded due to poor quality

sequencing as outlined in Methods.

Table 2-4: Poor quality variants excluded from burden analysis

Gene and variant DCM HVOL Total BAG3 1 1 c.636C>G 1 1 DES 1 1 c.1244G>A 1 1 DSC2 6 6 c.2048C>A 2 2 c.2056G>C 1 1 c.2059G>C 2 2 c.2071C>A 1 1 DSP 1 1 c.23A>C 1 1 FKTN 2 2 c.1261_1275delGCCAACTATGGTAAG 1 1 c.1280_1287delGGAAGATT 1 1 GATAD1 1 1 c.175_177delGGC 1 1 JUP 2 2 c.708-1G>C 1 1 c.708-2A>C 1 1 LDB3 3 4 7 c.1324G>C 2 2 c.1330G>C 2 1 3 c.1384A>G 1 1 2 MURC 7 4 11 c.702_722delAAGGCAGTCAGGAGAGAGGCT 7 4 11 MYBPC3 1 1 2 c.2858G>T 1 1 c.325G>T 1 1

130 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Gene and variant DCM HVOL Total PKP2 1 1 2 c.1460A>G 1 1 2 PRDM16 1 1 c.2780A>C 1 1 RBM20 3 1 4 c.166G>C 1 1 c.173_174delTA 3 3 SGCB 1 1 2 c.21_23delGGC 1 1 2 TTN 1 6 7 c.30683-1G>T 1 1 c.37461A>T 1 1 c.38066_38067insAGTGCCAGTGCCAGTGCCTCCTCCTAAAAAGCC 1 2 3 c.38067_38074delTGAAAAGA 1 1 c.38075_38076insGTGCC 1 1 VCL 11 11 c.152T>G 11 11

2.4.4 Burden of rare genetic variation in DCM patients compared to the

ExAC cohort

The overall excess burden of protein altering variants (truncating and non truncating) in

Caucasian DCM patients compared to Non-Finnish European ExAC cohorts is shown in

Figure 2-5.

Of the genes significantly enriched, variants in the TTN gene are the most common in this

DCM cohort, followed by non-truncating variants in MYH7 and TNNT2, and truncating

variants in DSP and LMNA (Figure 2-6). These are described further in the following

sections.

131 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Figure 2-5: Excess burden of protein altering variants (truncating and non-truncating) in DCM genes in

DCM cohort compared to ExAC. Titin variants are excluded in the right hand plot.

TTN non-tv, 7.0 MYH7 TTN tv, non-tv, 2.8 14.1 TNNT2 non-tv, 0.8 DSP tv, 0.9 Negative, Other, 17.6 56.4 LMNA tv, 0.4

Figure 2-6: Pie chart of excess burden (%) in significantly enriched genes (p<0.0009) in DCM cohort

compared to ExAC.

132 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.4.4.1 Truncating variants

Truncating variants in the genes encoding titin (TTNtv), desmoplakin (DSPtv), and lamin

A/C (LMNAtv) were significantly enriched in DCM cases compared to ExAC (Figure 2-7,

Table 2-5). Of these, TTNtv were the commonest variant in DCM cases, with an excess

burden of 14.1% (p=6.34 x10-82), followed by DSPtv (0.85%, p=2.5 x10-5), then LMNAtv

(0.4%, p=0.0006).

Figure 2-7: Burden of rare truncating variants in DCM compared to ExAC. Truncating variants in the

labelled red genes (TTN, DSP, and LMNA) are significantly enriched in DCM cases compared to ExAC

(p<0.0009). The right hand plot excludes truncating variants in titin (TTN). Bars indicate 95% binomial

confidence intervals, for clarity shown only for significant genes.

133 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Table 2-5: Burden of rare truncating variants in DCM genes in DCM cohort (Caucasian) compared to

ExAC (NFE). Burden testing performed using Fisher’s exact test. Bonferonni corrected significance

threshold p<0.0009.

Gene No of DCM DCM No of ExAC ExAC % excess p-value variants total frequency variants total frequency in DCM in DCM in ExAC compared cohort cohort to ExAC TTN 96 647 0.1484 253 32435 7.80E-03 14.1 6.34E-82 DSP 6 647 0.0093 25 32752 7.63E-04 0.85 2.52E-05 LMNA 3 647 0.0046 5 28041 1.78E-04 0.44 0.00059 ANKRD1 2 647 0.0031 13 33129 3.92E-04 0.27 0.03 FKTN 2 647 0.0031 13 33017 3.94E-04 0.27 0.03 TBX5 1 647 0.0015 1 32833 3.05E-05 0.15 0.04 PRDM16 1 647 0.0015 1 31468 3.18E-05 0.15 0.04 MYH7 2 647 0.0031 18 33249 5.41E-04 0.26 0.06 RAF1 1 647 0.0015 2 33285 6.01E-05 0.14 0.06 ABCC9 2 647 0.0031 27 33246 8.12E-04 0.23 0.11 TNNT2 1 647 0.0015 9 31444 2.86E-04 0.12 0.18 TXNRD2 1 647 0.0015 9 31372 2.87E-04 0.12 0.18 MYPN 1 647 0.0015 10 33272 3.01E-04 0.12 0.19 PKP2 1 647 0.0015 22 32105 6.85E-04 0.08 0.37 LAMA2 2 647 0.0031 100 33006 3.03E-03 0.01 0.59 DSC2 0 647 0 9 32997 2.73E-04 -0.03 1 DOLK 0 647 0 6 33251 1.80E-04 -0.02 1 MURC 0 647 0 18 32939 5.46E-04 -0.05 1 VCL 0 647 0 10 32637 3.06E-04 -0.03 1 DSG2 0 647 0 22 32937 6.68E-04 -0.07 1 TNNC1 0 647 0 1 32414 3.09E-05 0.00 1 ACTC1 0 647 0 1 33073 3.02E-05 0.00 1 CASQ2 0 647 0 14 32995 4.24E-04 -0.04 1 ILK 0 647 0 13 32827 3.96E-04 -0.04 1 JUP 0 647 0 3 30975 9.69E-05 -0.01 1 SGCB 0 647 0 6 33302 1.80E-04 -0.02 1 ACTA1 0 647 0 7 32721 2.14E-04 -0.02 1 ACTN2 0 647 0 4 33199 1.20E-04 -0.01 1 BAG3 0 647 0 3 32671 9.18E-05 -0.01 1

134 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Gene No of DCM DCM No of ExAC ExAC % excess p-value variants total frequency variants total frequency in DCM in DCM in ExAC compared cohort cohort to ExAC CRYAB 0 647 0 3 27907 1.07E-04 -0.01 1 DES 0 647 0 12 27928 4.30E-04 -0.04 1 DMD 0 647 0 10 31320 3.19E-04 -0.03 1 GATAD1 0 647 0 6 32686 1.84E-04 -0.02 1 MYL2 0 647 0 9 33279 2.70E-04 -0.03 1 NKX2-5 0 647 0 1 27528 3.63E-05 0.00 1 NPPA 0 647 0 6 33039 1.82E-04 -0.02 1 PLN 0 647 0 1 33247 3.01E-05 0.00 1 SGCD 0 647 0 3 27960 1.07E-04 -0.01 1 TAZ 0 647 0 1 27865 3.59E-05 0.00 1 TBX20 0 647 0 4 32038 1.25E-04 -0.01 1 TCAP 0 647 0 3 30370 9.88E-05 -0.01 1 TNNI3 0 647 0 14 29802 4.70E-04 -0.05 1 TPM1 0 647 0 1 32332 3.09E-05 0.00 1 ZBTB17 0 647 0 0 29982 0.00E+00 0.00 1 CSRP3 0 647 0 16 33337 4.80E-04 -0.05 1 DNAJC19 0 647 0 8 30955 2.58E-04 -0.03 1 LDB3 0 647 0 16 31891 5.02E-04 -0.05 1 MYBPC3 0 647 0 26 26194 9.93E-04 -0.10 1 SCN5A 0 647 0 16 31392 5.10E-04 -0.05 1 TMPO 0 647 0 19 32447 5.86E-04 -0.06 1 LAMA4 0 647 0 28 32844 8.53E-04 -0.09 1 PDLIM3 0 647 0 14 32285 4.34E-04 -0.04 1 NEXN 0 647 0 28 32704 8.56E-04 -0.09 1 MYH6 0 647 0 31 33165 9.35E-04 -0.09 1 EMD 0 647 0 0 32132 0.00E+00 0.00 NA LAMP2 0 647 0 0 32758 0.00E+00 0.00 NA

135 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.4.4.2 Non-truncating variants

Non-truncating variants in the genes MYH7 (2.8% excess, p=6.4 x10-7), TNNT2 (0.8%

excess, p=0.0003), and TTN (7.0% excess, p=8.3 x10-5) were significantly enriched in DCM

cases compared to ExAC (Figure 2-8, Table 2-6).

Figure 2-8: Burden of rare non-truncating variants in DCM compared to ExAC. Non-truncating variants

in the labelled red genes (MYH7, TNNT2, and TTN) are significantly enriched in DCM cases compared

to ExAC (p<0.0009). The right hand plot excludes variants in titin (TTN). Bars indicate 95% binomial

confidence intervals, for clarity shown only for significant genes.

136 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Table 2-6: Burden of rare non-truncating variants in DCM genes in DCM cohort (Caucasian) compared

to ExAC (NFE). Burden testing performed using Fisher’s exact test. Bonferonni corrected significance

threshold p<0.0009.

Gene No of DCM DCM No of ExAC ExAC % excess in p-value variants total frequency variants in total frequency DCM in DCM ExAC compared cohort cohort to ExAC MYH7 27 647 0.0417 449 33249 1.35E-02 2.82 6.38E-07 TTN 237 647 0.3663 9600 32435 2.96E-01 7.03 8.34E-05 TNNT2 6 647 0.0093 40 31444 1.27E-03 0.80 0.00031 PKP2 13 647 0.0201 274 32105 8.53E-03 1.16 0.005 DES 8 647 0.0124 110 27928 3.94E-03 0.85 0.0055 DSP 28 647 0.0433 832 32752 2.54E-02 1.79 0.0057 LAMA2 32 647 0.0495 998 33006 3.02E-02 1.93 0.0058 MYBPC3 19 647 0.0294 405 26194 1.55E-02 1.39 0.0076 NEXN 10 647 0.0155 218 32704 6.67E-03 0.88 0.01 ACTA1 3 647 0.0046 24 32721 7.33E-04 0.39 0.02 TNNC1 2 647 0.0031 12 32414 3.70E-04 0.27 0.03 LMNA 8 647 0.0124 153 28041 5.46E-03 0.69 0.03 PRDM16 16 647 0.0247 483 31468 1.53E-02 0.94 0.05 TXNRD2 7 647 0.0108 158 31372 5.04E-03 0.58 0.05 TPM1 2 647 0.0031 20 32332 6.19E-04 0.25 0.07 CRYAB 3 647 0.0046 39 27907 1.40E-03 0.32 0.07 SGCD 4 647 0.0062 65 27960 2.32E-03 0.39 0.07 TCAP 3 647 0.0046 56 30370 1.84E-03 0.28 0.13 ANKRD1 3 647 0.0046 72 33129 2.17E-03 0.24 0.17 NKX2-5 3 647 0.0046 62 27528 2.25E-03 0.23 0.19 TBX20 3 647 0.0046 76 32038 2.37E-03 0.22 0.21 SCN5A 14 647 0.0216 527 31392 1.68E-02 0.48 0.21 FKTN 4 647 0.0062 128 33017 3.88E-03 0.23 0.25 TNNI3 2 647 0.0031 47 29802 1.58E-03 0.15 0.28 DSC2 7 647 0.0108 278 32997 8.43E-03 0.24 0.31 DOLK 4 647 0.0062 147 33251 4.42E-03 0.18 0.33 MYPN 8 647 0.0124 350 33272 1.05E-02 0.19 0.38 DMD 14 647 0.0216 617 31320 1.97E-02 0.19 0.4 ABCC9 5 647 0.0077 215 33246 6.47E-03 0.12 0.41

137 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

Gene No of DCM DCM No of ExAC ExAC % excess in p-value variants total frequency variants in total frequency DCM in DCM ExAC compared cohort cohort to ExAC LAMA4 12 647 0.0185 561 32844 1.71E-02 0.14 0.43 TAZ 1 647 0.0015 27 27865 9.69E-04 0.05 0.47 GATAD1 1 647 0.0015 33 32686 1.01E-03 0.05 0.49 ILK 3 647 0.0046 134 32827 4.08E-03 0.05 0.5 ACTN2 6 647 0.0093 297 33199 8.95E-03 0.04 0.52 LDB3 5 647 0.0077 240 31891 7.53E-03 0.02 0.54 SGCB 2 647 0.0031 94 33302 2.82E-03 0.03 0.55 JUP 6 647 0.0093 297 30975 9.59E-03 -0.03 0.59 EMD 1 647 0.0015 44 32132 1.37E-03 0.01 0.59 MYL2 1 647 0.0015 46 33279 1.38E-03 0.01 0.6 LAMP2 1 647 0.0015 50 32758 1.53E-03 0.00 0.63 PDLIM3 2 647 0.0031 122 32285 3.78E-03 -0.07 0.7 ZBTB17 2 647 0.0031 127 29982 4.24E-03 -0.11 0.76 VCL 4 647 0.0062 263 32637 8.06E-03 -0.19 0.76 BAG3 3 647 0.0046 214 32671 6.55E-03 -0.20 0.8 MYH6 10 647 0.0155 644 33165 1.94E-02 -0.39 0.8 DSG2 4 647 0.0062 288 32937 8.74E-03 -0.25 0.82 CSRP3 1 647 0.0015 91 33337 2.73E-03 -0.12 0.83 RAF1 1 647 0.0015 91 33285 2.73E-03 -0.12 0.83 TMPO 2 647 0.0031 195 32447 6.01E-03 -0.29 0.9 MURC 1 647 0.0015 135 32939 4.10E-03 -0.26 0.93 TBX5 0 647 0 120 32833 3.65E-03 -0.37 1 NPPA 0 647 0 68 33039 2.06E-03 -0.21 1 DNAJC19 0 647 0 30 30955 9.69E-04 -0.10 1 PLN 0 647 0 6 33247 1.80E-04 -0.02 1 ACTC1 0 647 0 21 33073 6.35E-04 -0.06 1 CASQ2 0 647 0 151 32995 4.58E-03 -0.46 1

138 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.4.4.3 Variants in RBM20

Given the poor coverage of RBM20 in ExAC, I compared the burden of variants in RBM20 in

the DCM cohort versus the healthy volunteer cohort. There was no evidence of a significant

difference in the burden of truncating or non-truncating variants in RBM20 between 647

DCM cases and 588 Caucasian healthy volunteers (Table 2-7).

Table 2-7: Burden of rare variants in RBM20 in DCM cohort compared to volunteers. Burden testing

performed using Fisher’s exact test. HVOL= healthy volunteer.

RBM20 DCM DCM HVOL HVOL Burden P value

Variant type variants frequency variants frequency DCM vs

N=647 N=588 HVOL (%)

Non-truncating 14 0.022 10 0.017 0.46 0.35

Truncating 2 0.003 1 0.0017 0.14 0.54

2.4.5 Burden of rare genetic variation in healthy volunteers compared to

the ExAC cohort

The ExAC dataset is not a control cohort but it is not expected to be enriched for DCM cases.

However, there is potential for there to be differences in the sensitivity of targeted panel

sequencing compared to whole exome sequencing for the target genes. Therefore the

difference between protein altering variants in ExAC and a cohort of 588 Caucasian healthy

volunteers who underwent sequencing using the same assay and platform as the DCM

patients was assessed.

There was a significant excess in protein altering variants in the genes TTN and NKX2-5 in

controls compared to ExAC, suggesting that the sequencing panel is more sensitive than

139 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

exome sequencing for these genes (Figure 2-9). Specifically with regards to TTN, there was

an excess of non-truncating variants in TTN, with an excess burden of 8.8% in volunteers

compared to ExAC (p=0.0004, Table 2-8). This suggests that the enrichment of TTN

missense variants in DCM compared to ExAC could have been driven by sensitivity of panel

sequencing, therefore I compared the burden of non-truncating TTN variants in DCM cases

and healthy controls, both subject to the same sequencing assay and platform. This showed

that there was no evidence of a significant difference in the burden of non-truncating variants

in TTN between DCM cases and healthy volunteers when assessed using these technically

matched cohorts (0.4% excess in DCM, p=0.004, p threshold for significance <0.0009, Table

2-8). Notably, the observed burden of TTNtv remained when controlling for coverage and

sensitivity differences using these cohorts. There was a significant excess of truncating

variants in TTN in DCM patients compared to healthy controls (14.3% excess in DCM, p=6.5

x10-25, Table 2-8).

Table 2-8: Burden of truncating and non-truncating variants in TTN in DCM patients compared to

ExAC and healthy volunteers. Burden testing performed using Fisher’s exact test.

Truncating Non-Truncating ExAC total 32435 32435 No of variants in ExAC 253 9600 HVOL total 588 588 No of variants in HVOLs 3 213 DCM total 647 647 No of variants in DCM 96 237 Excess burden DCM vs HVOL (%) 14.3 0.41 p-value DCM vs HVOL 6.54E-25 0.46 Excess burden DCM vs ExAC (%) 14.1 9.2 p-value DCM vs ExAC 6.34E-82 8.34E-05 Excess burden HVOL vs ExAC (%) -0.2 8.8 p-value HVOL vs ExAC 0.83 0.0004

140 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

The apparent enrichment in NKXK2-5 may also be compounded by the relatively low

coverage of this gene in ExAC, with only 67% of sample bases covered ≥10x.

There was no significant difference in the frequency of variation in the other genes in

controls compared to ExAC, including the genes LMNA, DSP, MYH7 and TNNT2 found to be

enriched in DCM compared to ExAC (Figure 2-9).

Figure 2-9: Burden of rare protein altering variants in volunteers compared to ExAC. Variants in the

labelled red genes (TTN and NKX2-5) are significantly enriched in healthy volunteers compared to ExAC

(p<0.0009). Bars indicate 95% binomial confidence intervals, for clarity shown only for significant genes.

HVOL= healthy volunteers.

141 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.4.6 Proportion of DCM cases with identifiable genetic variant

The proportion of individuals with a genetic variant in the five genes significantly enriched in

DCM compared to background population variation (TTNtv, DSPtv, LMNAtv, MYH7ms,

TNNT2ms) is shown in Figure 2-10. Amongst patients with a family history of DCM (n=92,

14%), a genetic variant is identified in 36% (n=33). Amongst patients without a family

history of DCM (n=555, 86%), a genetic variant is identified in 18% (n=101) (p=0.0003).

Of those patients with genetic variants, the majority had only one genetic variant in the target

genes. One patient with a TTNtv had a missense variant in MYH7. No other patients had two

variants in the target genes. Overall, in this cohort, 79% of cases with DCM remain

genetically unexplained.

Figure 2-10: Bar chart showing percentage of DCM cases with identified genetic variant in TTN, LMNA,

DSP, MYH7 and TNNT2 (variant classes in text). Cases are stratified by a family history of DCM.

142 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.5 Discussion

This study, evaluating the genetic architecture of DCM, shows that truncating variants in titin

are the commonest genetic cause of DCM, with an excess burden of ~14% in Caucasian

DCM patients. Many previous studies that have reported genetic association with DCM have

not fully controlled for the high level of rare genetic variation in the population. This is the

first large scale study in DCM to do so. In this cohort, variation in only a further 4 genes

(LMNA, DSP, MYH7, TNNT2) are significantly enriched in DCM compared to background

population variation. This is much lower than previous studies have suggested and the

reasons for this are explored further. This study also demonstrates that a large proportion of

DCM cases (79%) remain genetically unexplained, even amongst those with a family history

of DCM (64%).

Frequency of TTNtv DCM – comparison to previously published cohorts

We identify that TTNtv in DCM represent the largest burden of genetic variation in DCM,

with a ~14% excess. The burden in the remaining 4 significantly enriched genes collectively

amounts to ~5%. Therefore, phenotype analysis in Chapter 3 will focus on TTNtv DCM.

Whilst the finding that TTNtv are the commonest genetic variant in DCM is in line with

previous reviews of studies in the field(5,67,308-310), the prevalence in this cohort differs

from previously published cohorts. In the landmark 2012 paper by Herman et al, confirming

the association between TTNtv and DCM across 3 clinical cohorts(79), the frequency of

TTNtv was reported as 25% in familial cases of idiopathic dilated cardiomyopathy and 18%

in sporadic cases. However, the frequency of TTNtv actually ranged from ~8% to 40% across

the cohorts, with the lowest frequency reported in a familial DCM cohort from Italy and

Colorado (13/149, 8.7%), sequenced using traditional sequencing methods. The higher

143 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

frequencies were reported in subjects recruited during evaluation for cardiac transplantation

(17/71, 23.9%) and severely affected patients from a specialist centre, with a mean LVEF of

26% (37/92, 40.2%).

The patients with TTNtv DCM in these latter 2 groups had a much higher event rate for

cardiovascular mortality, transplant or ventricular assist device implantation (100% and 38%

respectively) compared to the first familial group (8%). Therefore it appears plausible that

TTNtv are enriched in severe/end stage cohorts, and a prevalence of 14% is a more accurate

reflection of an unselected DCM population. This hypothesis is supported by the largest

sequencing study of DCM patients referred for clinical genetic testing to date, in which 766

DCM patients underwent clinical genetic sequencing(67). Amongst those sequenced for TTN

(up to 25% of the cohort; inconsistencies reflected the duration of the study in which newer

genes were not sequenced in all patients), overall, TTNtv were found in 14% of cases(311).

Similarly, in a previous study of TTNtv in DCM patients from our institution by Roberts et

al(84), the prevalence of TTNtv was 13% in unselected DCM cases (49/374) and 22% in an

end-stage cohort (34/155), the latter overlapping with the transplant cohort in the original

Herman paper. Of note, 236 patients within the unselected cohort reported by Roberts et al

are included within the DCM cohort in this chapter. Finally, in a study of 72 DCM probands

with severely impaired ventricular function (mean LVEF 24%), TTNtv were identified in

24% of cases (30% of familial DCM, 18% of sporadic DCM)(86).

DCM genetic architecture – Fewer genes contribute to genetic architecture of DCM

than previously estimated

The second key finding of this study is that in this cohort, of the 57 genes reported to be

associated with DCM, only 5 were significantly enriched in cases compared to background

144 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

population genetic variation. This suggests that previous estimates of the genetic architecture

of DCM may have been conflated by inadequate control for background population variation.

Indeed, the large scale exome sequencing project ExAC (Exome Aggregation Consortium)

has demonstrated a relatively high level of rare coding variants at a population level, with

approximately 54 variants per participant previously reported as disease causing(15).

With respect to DCM, the significantly enriched variants were TTNtv as outlined, as well as

truncating variants in DSP and LMNA, and non-truncating variants in MYH7 and TNNT2.

Non-truncating variants in TTN (TTNms) were not enriched in DCM compared to ExAC

when the sensitivity of sequencing panel was accounted for, as healthy volunteers in our

cohort also demonstrated an excess of TTNms variants compared to ExAC.

The findings of this Chapter are primarily used to inform phenotype analysis in subsequent

chapters. However, they also inform us about the genetic architecture of DCM and this data

may inform future patient stratification studies. This is the largest study to date of DCM

patients sequenced on a comprehensive targeted sequencing panel (including 57 DCM

genes). A demonstrated excess burden in a particular gene therefore provides good evidence

that the gene in question is a valid DCM gene. However, the significance of the absence of

excess for the remaining 52 genes requires a more nuanced approach. There are at least two

key possibilities.

Understanding the genes that were not enriched in DCM cases

The first possibility for a lack of enrichment is that the gene in question is not a valid DCM

gene, and the original reports either did not fully account for population variation, did not

provide adequate segregation data or did not provide adequate functional data. Indeed, many

145 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

of the variants in genes linked to DCM were reported in relatively small case-control cohorts,

prior to our current knowledge of the extent of rare background genetic variation, and in the

absence of formal statistical association testing. Many report simply the presence of variants

in a case cohort and the absence of the specific variant in a similar sized control cohort. For

example the first reported link between MYBPC3 and DCM identified one missense variant

in 46 patients with DCM, absent in 88 healthy controls and 136 patients with hypertrophic

cardiomyopathy(44). Others report the absence of a particular variant in large datasets, but do

not consider any variation across the gene, for example the study linking PRDM16 to DCM,

in which 206 samples from DCM patients (including 131 cardiac biopsies) were sequenced.

In total, 2 truncating (0.97%) and 5 missense variants (2.4%) were identified which were

each absent in 6,400 controls. However, the authors did not report the total rare variation in

PRDM16 in this series. In the ExAC NFE cohort, 483 non-truncating variants (1.4%) and 1

truncating variant (0.004%) in PRDM16 are found in 31,468 samples(58). In the original

study, the cases with missense variants lacked segregation data and two of the missense

variants were annotated in dbSNP, with predicted benign PolyPhen scores(59). The two de-

novo truncating variants may be of interest and indeed, in our cohort, one DCM patient

carried a truncating variant in PRDM16 (burden testing compared to ExAC, p=0.002).

However, based on the current proportions, to evaluate this, a sample size of over 4000

patients with DCM would be required (calculation performed using power.prop.test in

R).

We purposely did not perform case-control burden analysis in this cohort, despite having 647

DCM cases and 588 controls, sequenced on the same platform using the same assay. Whilst

588 controls is a large sample size, it is possibly insufficient to be able to adequately capture

rare population genetic variation. For example, if we consider non-truncating variants in the

146 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

gene ANKRD1. These were first reported to be associated with DCM in a study of 208 DCM

patients (312). The authors identified 3 missense mutations in ANKRD1 in 4 DCM patients,

absent in 180 control subjects in the dbSNP database. In our cohort, 3 DCM patients (0.5%)

have missense variants in ANKRD1 and there are no variants in the control subjects.

However, there are 72 missense variants in ANKRD1 in 33129 NFE subjects in ExAC (0.3%)

and no significant enrichment in cases compared to ExAC (p=0.17) or controls (p=0.14) on

burden testing.

The second possibility is that the gene in question is a valid DCM gene, but only for variation

in particular domains or hotspots, particularly when the original reports suggested that such

hotspots existed, for example variants in RBM20(55). As outlined, we were not able to fully

evaluate variations in RBM20 due to its low coverage in ExAC. This issue may be addressed

in future studies evaluating the genetic architecture of DCM compared to population variation

using whole genome sequencing data in gnomAD (Genome Aggregation Database)(313),

which appears to have a better coverage of RBM20 compared to ExAC. Finally, there may be

particular variants within a gene that are pathogenic, especially in the context of a Founder

mutation, for example PLN R14del(65), which will not be accounted for in the study design

of this project.

Other possible explanations for the genes that were not enriched could also be that whilst the

gene is a true disease gene, mutations are rare and account for a very small proportion of

cases, much lower than the threshold we are powered to detect. Furthermore, the assumption

tested in this study is that variation in each individual gene is sufficient to cause disease. If

their contribution however is more complex than monogenic, this will not be detected in this

study design.

147 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

In this study, variants in the remaining 52 genes without strong evidence for pathogenicity,

according to the ACMG guidelines(36), may still be of importance for the individual patient,

but should only be classified as pathogenic with robust additional data, for example

segregation data. Moving forwards, large scale collaborative efforts (such as that proposed by

ClinGen) will be required to comprehensively review the available evidence linking the 57

genes to DCM to assess whether they should be incorporated in clinical genetic testing.

Previous studies evaluating the genetic architecture of DCM

I now place the result of this study in the context of earlier studies evaluating the genetic

architecture of DCM. The first, by Pugh et al, 2014, analysed variants in up to 46 cardiac

genes in 766 DCM patients referred for clinical genetic testing(67). Classifying variants as

pathogenic or likely pathogenic, using NHLI-ESP and 1000 Genomes to account for

population variation, they identified that TTNtv were the most common pathogenic variant

(12.2%), followed by variants in LMNA (4.1%), MYH7 (3.4%), TNNT2 (2.5%) and DSP

(2.4%), very much in line with our results. Importantly, the authors did not detect any

“pathogenic”, “likely pathogenic”, or “VUS (Variants of uncertain significance)-favour

pathogenic” variants in 24 of the 46 genes tested, including 11 which had previously been

reported to be associated with DCM (ANKRD1, CAV3, CRYAB, CTF1, DSC2, EMD, FHL2,

LAMA4, LAMP2, MYH6, and PKP2). Furthermore, the authors identified only 3 “VUS-

favour pathogenic” variants in the gene MYBPC3 across the cohort, 2 of which may have

been variants more associated with hypertrophy cardiomyopathy, reinforcing our finding that

MYBPC3 is not a DCM gene.

In the second study, published by Haas et al in 2015, 639 DCM patients from 8 countries

underwent targeted sequencing of 84 genes(314). No control populations were sequenced.

148 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

The findings of the study by Haas et al are distinct to our study. Initially, the authors report

that 48% of patients have variants in known cardiomyopathy genes. However, when

restricted to variants that were absent in NHLI-ESP and included variants in DCM genes

only, variants were only detected in 16% of patients. The estimated frequency of genetic

variants was higher than in any previous study, with 73% of patients reported to have a

pathogenic variant and strikingly 38% of patients having at least two pathogenic variants.

However, this was likely related to the approach to variant classification that the authors

used. In contrast to ACMG guidelines(36), the authors adopted the following approach to

variant classification.

The primary inclusion category was variants in HGMD, including those linked to all

cardiomyopathies and channelopathies (not specifically DCM). The secondary categories

were ‘potential disease causing variants’, considered separately for truncating and non-

truncating variants, the latter annotated with a SNP annotation tool. Crucially, the frequency

threshold for exclusion was not clear, with the authors defining ‘not common’ variants.

Historically, rare variants are typically defined as having a minor allele frequency (MAF)

<1%, though the frequency cut offs in the literature vary(23). For SNPs, the authors excluded

variants with a frequency over 1%, but work from our group has shown that this threshold is

high and is likely to include many benign variants(14,24). The maximum credible population

allele frequency for any rare variant causative of DCM is <0.00006 (calculation outlined in

Section 3.3.3.2). Therefore, many of the variants in the study by Haas et al may have been

variants that are well tolerated at a population level. Re-evaluation of the original data with

reference to either the ExAC or gnomAD datasets would be of great interest.

149 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

2.5.1 Limitations of this study

There are a number of potential limitations to this work. Firstly, I am comparing the burden

of variants detected through two distinct sequencing technologies, namely targeted

sequencing versus whole exome data. This means the two cohorts are not uniformly matched

and there is potential for bias, with differences detected being secondary to the sequencing

method used. The coverage in whole exome data, particularly for the targeted genes, is not as

high as in targeted sequencing data. I have taken steps to mitigate this, including evaluating

the frequency of the variant as a function of the actual number of individuals sequenced for

that gene in ExAC and not the total number of individuals in ExAC, though this remains a

potential limitation. However, whilst there are clear limitations, the strength of using the

ExAC dataset is for its size, enabling the detection of rare variants that would not be possible

in smaller matched case-control cohorts. We did also recruit and sequence an ethnicity

matched control cohort, sequenced on the same assay and platform, therefore were able to use

this data to evaluate the sensitivity of the panel compared to ExAC. This analysis showed that

the targeted panel was more sensitive to missense variation in TTN and NKXK2-5 compared

to exome sequencing in ExAC. For NKXK2-5, the sensitive panel did not lead to false

positive enrichment in cases. There was no evidence to support NKXK2-5 being a DCM gene

as there was no excess variation in DCM cases compared to controls or in DCM cases

compared to ExAC. With regards TTN missense variants, whilst these were enriched in DCM

cases compared to ExAC, this is likely driven by the sensitivity of the targeted panel. TTN

missense variants may still be of importance in DCM but further work is required to

determine which variants may act as disease modifiers and which are innocent bystanders.

As reviewed in the discussion, having both the control cohort and ExAC shows that reliance

on the control cohort alone for burden comparison may have been insufficient. Together,

150 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

these limitations would have potentially led to an increase in false positive disease-gene

associations. On the contrary, the findings of this study do not support many of the reported

DCM disease-gene associations, but reinforce an emerging literature that the number of genes

causally implicated in monogenic DCM is smaller than previously estimated.

Due to the primary analysis being a burden comparison between targeted sequencing data and

whole-exome data, burden testing was performed using Fisher’s test, based on the collapsed

and combined count frequencies of variants in each cohort. This gene based burden testing

may be associated with reduced power compared to other statistical methods, particularly if

the rare variants may increase or decrease disease risk(315), though for DCM, we expect the

rare variants to increase disease risk. The factors affecting the power of this method of gene-

based burden testing using population data have recently been comprehensively

reviewed(316). Using model simulations on a set of ~2,600 whole exome sequencing

samples, the authors found that the primary determinants of power in such studies are the

background rates of rare variants in each gene, with genes such as TTN, with a high

background rate of variation, requiring large samples (approaching 5,000) to detect a

difference between cases and controls(316). The second key determinant of power was locus

heterogeneity, with larger sample sizes needed to achieve adequate power when individual

genes contribute to less than 5% of cases under a dominant model, as is the case with DCM.

However, for the lowest estimate (contribution of any individual gene= 1%), just over 300

case samples were required to achieve 80% power to detect a difference between case and

population cohorts, a sample size achieved in this study. Surprisingly, incomplete penetrance,

even as low as 10%, did not affect power. This was attributed to the overall low prevalence of

monogenic disorders(316). In our study, based on the simulated power calculations, we are

powered at 80% to detect a significant enrichment of genes with an excess burden of 2.7% in

151 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

cases compared to population control. This means that it is unlikely that the significantly

enriched genes with an excess burden lower than this threshold are false positives (e.g.

TNNT2 non-truncating variants). It also however remains plausible that many of the lower

frequency genes that were not significantly enriched may be re-evaluated in future studies

with larger case sample size.

Another limitation of this study design and analysis is that non-coding variants in the genes

evaluated or the effects of epigenetic modifications, which may be important contributors to

DCM pathogenesis are not accounted for. Similarly, gene-gene interactions, structural

variants, or any common variants of small effect size are also not accounted for. Future

studies evaluating gene burden based on whole genome sequencing of DCM patients will be

important both for these reasons, as well as mitigating the bias associated with incomplete

coverage of the coding region in targeted sequencing and exome sequencing due to GC-rich

regions(317,318).

Steps have been taken to match the case and control cohorts for ethnicity to avoid false

positive gene-disease associations. However, this was based on self reported ethnicity and not

principal component analysis (PCA) due to the limited number of targeted SNPs in the panel

sequencing data. Work is now in development within our group to evaluate options to be able

to perform SNP based PCA on data from the TruSight Cardio panel. Future gene-based

burden studies will therefore be able to stratify by PCA-derived ethnicity in this cohort.

The results in this study could also be affected by selection bias within the primary study

cohort. The majority of the cohort underwent CMR at recruitment, which means that there is

potential bias against either individuals with severe disease (who were for example too

152 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

unwell to undergo CMR), or individuals with non-MRI conditional or compatible cardiac

devices. The latter is particularly important as LMNA cardiomyopathy is known to be

arrhythmogenic and patients with this condition may have pacemakers or defibrillators

implanted(49). This may explain the relatively lower frequency of LMNA variants in this

cohort compared to previously published estimates up to 5%(5), as well as the absence of

significant enrichment of non-truncating variants in LMNA in DCM cases compared to

population controls.

The study cohort is also predominantly composed of individuals with adult onset DCM, with

a mean age at recruitment of 54 years. This means that we are not enriched for cases of

paediatric DCM, which is known to have a different genetic architecture to adult onset DCM,

both in terms of genes involved (e.g. TPM1 is associated with paediatric DCM) as well as

mode of inheritance (more evidence of recessive models)(319). This was illustrated nicely in

the series by Pugh et al, in which different gene burdens were identified when the primary

population was stratified by adult or paediatric status(67). To mitigate these ascertainment

biases, we have now identified a secondary replication cohort composed of patients with a

history of DCM referred for clinical genetic testing, both from the Partners Laboratory of

Molecular Medicine(67) and Oxford Medical Genetics Laboratory.

2.6 Summary

In this chapter, the genetic architecture of DCM was re-evaluated in the context of population

level genetic variation. This identified that truncating variants in TTN are the largest genetic

contributor to DCM with an excess of ~14% in Caucasian patients. Variants in only a further

4 genes (LMNA, DSP, MYH7, TNNT2) were significantly enriched in DCM compared to

153 Template by Friedman & Morgan 2014 Word Chapter 2: Evaluating the genetic architecture of dilated cardiomyopathy

background population variation, leaving ~79% of DCM cases in this cohort genetically

unexplained. This data will direct the genotype-phenotype analysis in subsequent chapters.

2.7 Outline of further work

• We are extending the analysis to a secondary replication cohort composed of patients

with a history of DCM referred for clinical genetic testing from the Partners

Laboratory of Molecular Medicine(67) and the Oxford Medical Genetics Laboratory.

• We are extending the analysis to non-Caucasian patients, both from the UK and a

secondary cohort of DCM patients sequenced in Singapore.

• The next phase of this study will replicate the analysis but using gnomAD as the

reference database, which will address some of the limitations of coverage

encountered when using the ExAC dataset.

2.8 Acknowledgements

• Roddy Walsh, senior Bioinformatician in Prof Cook’s group, downloaded the ExAC

dataset, extracted the list of variants in the target genes from the ExAC dataset, the

DCM cohort, and the control cohort, and summarised the data in a format suitable for

burden analysis.

• Risha Govind, Bioinformatician in Prof Cook’s group, provided the coverage data for

the targeted sequencing panel samples, and together with Shibu John ran the

bioinformatics pipeline.

• Rachel Buchan, William Midwinter, and Alicja Wilk, laboratory scientists in Prof

Cook’s group, performed the sequencing of the DCM and HVOL samples.

154 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3 THE CLINICAL

MANIFESTATIONS AND

PHENOTYPIC DRIVERS OF TITIN

CARDIOMYOPATHY

3.1 Aims and Hypotheses

The primary aim of this chapter is to define the cardiovascular phenotype of titin

cardiomyopathy in patients with DCM, with the purpose of informing patient stratification

and identifying novel insights into disease pathogenesis. The hypotheses are as outlined:

- The cardiovascular phenotype of dilated cardiomyopathy in patients with titin truncating

variants is distinct to that observed in patients without titin truncating variants.

- Alcohol consumption is an environmental modifier of the cardiovascular endophenotype

of left ventricular ejection fraction in patients with DCM and truncating variants in titin.

3.2 Background

As previously outlined, despite improvements in pharmacological and device based therapy,

155 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

clinical outcomes in DCM remain poor, with a 20% 5-year mortality rate (12,13). As such

there is a pressing need for improved risk stratification and the identification of novel

therapeutic targets. As presented in the previous chapter, truncating variants in the gene for

the sarcomeric protein titin (TTNtv) are the largest single genetic contributor to DCM, found

in up to 14% of cases in this cohort, and 10-20% of cases in previously reported cohorts

(79,84). Identifying genotype-phenotype correlations for TTNtv DCM may therefore

advance stratification approaches and treatment options for DCM patients.

In this chapter, the deep phenotyping attributes of cardiovascular magnetic resonance (CMR)

are used to gain insights into the pathobiology of TTNtv DCM. The identification of a

distinct cardiac endophenotype of TTNtv DCM may offer mechanistic insights into the

pathogenesis of DCM and highlight potential therapeutic targets for further translational

study.

There are key unanswered questions with respect to titin cardiomyopathy, which I set out to

explore in this chapter:

1. Genotype – phenotype correlations: As yet, there are few clear genotype-phenotype

correlations for TTNtv DCM. If they can be established, they may form the basis of

improved risk stratification in DCM.

2. Variable penetrance and expressivity: TTNtv are notable for variable penetrance and

expressivity. I explore the role of environmental modifiers, with a focus on alcohol

consumption.

3. Mechanism of disease: The mechanism of action of TTNtv leading to DCM is

incompletely understood. Studying the human phenotype may provide mechanistic

insight and identify novel therapeutic targets.

156 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Environmental modifiers

Informed genotype-phenotype risk stratification requires an understanding of potential

phenotypic modifiers. Genetic or environmental modifiers may underpin the variable

penetrance and expressivity seen in TTNtv DCM. Indeed, there is now good evidence for

environmental modifiers of the TTNtv phenotype. For example, peripartum cardiomyopathy

shares a genetic aetiology with DCM, suggesting that pregnancy may act as an environmental

modifier to unmask the phenotype of cardiomyopathy in TTNtv carriers (92). We have also

recently demonstrated that TTNtv are not phenotypically silent in otherwise apparently

healthy humans, with subtle differences in cardiac volumes and physiology compared to

control subjects without TTNtv (88). Furthermore, we found that rats with TTNtv developed

impaired cardiac physiology under cardiac stress (88), again providing evidence of the

potential role of gene-environmental interactions in the development of the TTNtv

cardiomyopathy phenotype.

Chronic excess alcohol consumption is an established contributor to cardiovascular

disease(96) and is known to cause cardiomyopathy (97). The effects of moderate alcohol

consumption however are less clearly established and the spectrum of alcoholic end organ

damage is not always related to quantitative consumption, suggestive of a putative genetic

modifier. Against this background, I therefore evaluate alcohol consumption as a phenotypic

modifier of TTNtv cardiomyopathy.

157 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.3 Methods

3.3.1 Study Population

The primary DCM study population, as outlined in Chapter 2, was prospectively enrolled in

the National Institute for Health Research (NIHR) Royal Brompton Hospital Cardiovascular

Biobank project between 2009-2015. This cohort consisted of consecutive referrals from a

dedicated cardiomyopathy service at the Royal Brompton Hospital, London, and consecutive

referrals to the CMR unit, both from within the Royal Brompton Hospital and a network of

thirty regional hospitals. Patients were referred for diagnostic evaluation, screening and

assessment of severity of DCM. Within the cardiomyopathy service, our standard protocol is

that all patients without contraindication undergo a CMR scan at presentation.

The control participants included for illustrative purposes only in Figure 3-4 are taken from a

previously published cohort from our institution(116). They consist of 120 healthy

volunteers, with 10 men and 10 women in each of six age deciles from 20 to 80 years. All

subjects were asymptomatic, had no contraindication to CMR scan, no history of cardiac

disease or known cardiac disease risk factors, and normal physical examination and ECG.

3.3.2 Ethics

All patients provided written informed consent. The study was approved by the regional

ethics committee (REC reference 09/H0504/104).

3.3.3 Titin truncating variant curation for phenotype analysis

Based on the results of Chapter 2, truncating variants in titin were the most enriched variant

in DCM compared to population reference samples and are the focus of the genotype-

158 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

phenotype analysis.

Genetic sequencing and bioinformatics analysis was performed as outlined in Chapter 2.

For the majority of patients, this was with the Illumina NextSeq platform and a customised

panel of 174 genes (TruSight Cardio, Illumina) as previously outlined. A small subset of

patients were sequenced on different platforms/assays. As part of the iterative process of

optimising sequencing metrics, sequencing was initially done on the SOLiD 5500 platform

(Life Technologies), and subsequently the MiSeq and HiSeq platforms (Illumina), with a

customised 201 gene panel assay, named Inherited Cardiac Conditions versions 4-6 (ICC

Sure Select, Agilent) from which the TruSight Cardio was developed. Gene panels are

summarised in Appendix Tables 1 and 2.

TTNtv were curated as outlined below. Truncating variants are defined as frameshift,

nonsense, and variants affecting the splice donor and acceptor regions (first and last two

bases of each intron).

3.3.3.1 Outline of variant filtering metrics

1. Using annotated VCF (variant call format) files, scripts were written to filter protein

altering variants in core DCM genes (Table 3-1) based on frequency and sequencing

metrics:

a. Minor allele frequency of variant in ExAC <0.0001* (threshold explained in

text)

b. Sequencing metrics:

i. Minimum coverage ≥ 10

1. Meaning at the variant position, each base has been sequenced

at least 10 times.

159 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

ii. QD score > 4

1. GATK specific filter. QualbyDepth (QD) refers to the variant

call confidence normalized by depth of sample reads supporting

a variant.

iii. Allelic balance ≥ 20

1. Allelic balance in heterozygous genotypes refers to the

proportion of reads arising from the reference and alternate

alleles and is a useful indicator of quality.

2. For TTNtv only:

a. Variants in constitutively expressed exons (percentage spliced in, PSI > 90%.

PSI explained in text).

b. Confirmation of variants through Sanger sequencing or review mapped

sequencing reads on Integrative Genomics Viewer(320,321).

Table 3-1: Genes and variant classes selected for variant curation. Only truncating variants (frameshift,

nonsense, essential splice site) in TTN, DSP, and VCL were evaluated. Only non-truncating variants (non-

synonymous SNP) were evaluated in MYH7 and TNNT2. For the remaining genes, both variant classes

were evaluated. This gene list was based on existing literature reports and the results from Chapter 2.

Gene Truncating Non-truncating Titin (TTN) + - Lamin (LMNA) + + Myosin heavy chain (MYH7) - + Troponin T2 (TNNT2) - + Desmoplakin (DSP) + - BAG3 + + RBM20 + + Tropomysin (TPM1) + + Vinculin (VCL) + -

160 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.3.3.2 DCM specific rare variant frequency threshold

Rare variants are typically defined as having a minor allele frequency (MAF) <1%, though

the frequency cut offs in the literature vary(23). However, for evaluation of potentially

pathogenic variants, a disease-specific frequency threshold for rare variants of <0.0001 was

used, informed by disease prevalence, penetrance, and allelic contribution to disease as

outlined below(24,25).

Calculating the maximum credible population allele frequency for any rare DCM

variant

!"#$%&% !"#$%&'# !"!#$%&'"( !""#"# !"#$%#&'( (!"#$%) =

!"#$%#$ !"#$%&#'(# × !"#$%&% !"#"$%& !" !""#"$% !"#$%&'($&"#

!"#$%#$ !"#"$%&#'"

Which can be rewritten as:

!"#$% = !"#$%&#'(# !"# !"#!$!#%&' × 0.5

!"v!"!#$ !"#$%&#'(# !"# !"#!$!#%&' !" !ℎ! !"#$%& !" !ℎ!"#"$"#%$ !"# !"#!$!#%&' ×

!"#$%#&'( !" !"##"$%&' !"#$"%& × ! !"#"$%&#'"

The threshold calculation as applied to DCM:

• Estimated disease prevalence 1/250(5)

• Commonest variant in a DCM cohort previously evaluated by our group:

TNNT2:c.629_631delAGA, 1.4%(14)

• Penetrance estimated to be 50%

=1/250 * 0.5 *0.014*(1/0.5) = 0.000056

161 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

This would therefore be the estimated maximum credible population allele frequency for any

rare variant causative of DCM. I have rounded up this frequency, therefore any variants

found at a population frequency of >0.0001 are unlikely to be pathogenic disease causing

variants in DCM.

3.3.3.3 Constitutively expressed exons in the titin gene

The TTN gene undergoes extensive alternative splicing resulting in the production of different

protein isoforms, meaning that not all exons are included in the final processed mRNA

transcripts. Allowing for this process, which is quantified by assessing the percentage spliced

in (PSI), that is the percentage of final cardiac transcripts that include a particular exon,

appears to be important for distinguishing variants that are important for disease. The PSI

data were derived during previous work from our group using tissue from patients with

DCM. This work demonstrated that variants in exons that are included in the final cardiac

transcript more than 90% of the time are most significant for human cardiomyopathy and are

associated with DCM with an odds ratio over 50 (84,322). These are referred to as

constitutively expressed exons. While overall TTNtv are not uncommon in the population, a

much smaller proportion of variants meet these criteria, with a prevalence of 0.004 (0.001 in

the A band)(322).

All TTNtv variants were annotated to the titin meta-transcript, manually curated by the

Havana Group (Table 3-2).

162 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-2: Titin meta-transcript details

Meta transcript

Description Inferred complete meta-transcript

Translation length (amino acids) 35991

Number of exons 363

Ensembl transcript ENST00000589042

Ensembl protein ENSP00000467141

Havana transcript: OTTHUMT00000450680

3.3.4 DCM phenotyping

3.3.4.1 Confirmation of DCM diagnosis in the Cardiovascular Biobank

Upon recruitment to the Biobank, participants were assigned a preliminary clinical diagnosis

by the recruiting research nurse. This was then reviewed by myself (from 2014 onwards) or

the Senior Investigators (Prof Cook or Dr Prasad, pre 2014) and participants were assigned

the diagnosis of DCM if an imaging diagnosis of dilated cardiomyopathy was confirmed,

either by echocardiography or cardiovascular magnetic resonance (CMR).

For patients with CMR, DCM was diagnosed based on evidence of left ventricular dilation

and systolic impairment with reference to age, gender, and body surface area adjusted

nomograms(116). For patients with echocardiography, DCM was diagnosed in the presence

of left ventricular end-diastolic diameter (LVEDd) > 117% of that predicted for age and body

surface and left ventricular ejection fraction (LVEF) < 45% and/or fractional shortening (FS)

< 25%(323).

163 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Analysis was restricted to patients with MRI data and this cohort was curated further to

ensure criteria for both an imaging and clinical diagnosis of DCM were met. This involved

review of MRI images, clinical records, patient questionnaire, family pedigree data and

ECGs. The exclusion criteria were a history of uncontrolled systemic hypertension, coronary

artery disease (>50% stenosis in one or more major epicardial arteries or previous

percutaneous coronary intervention or coronary artery bypass grafting), chronic excess

alcohol consumption meeting criteria for alcoholic cardiomyopathy (>80g/day for more than

5 years(97)), systemic disease known to cause DCM, pericardial disease, congenital heart

disease, infiltrative disorders (e.g. sarcoidosis), or significant primary valvular

disease(8,306,307).

3.3.4.2 Cardiovascular magnetic resonance

Patients in the CMR cohort underwent CMR at 1.5T (Siemens Sonata or Avanto scanners).

Breath hold steady state free precession cine images were acquired in 3 long axis planes and

successive 8mm short axis slices (2mm gap) from the atrioventricular ring to the apex (13).

Late gadolinium enhanced (LGE) images were acquired using a breath hold inversion

recovery sequence following administration of 0.1mmol/kg of gadolinium contrast agent

(Magnevist or Gadovist, Bayer), with inversion times optimised to null normal myocardium.

Mid-wall myocardial fibrosis was recorded as present if detected in the primary and phase

swapped image and in 2 orthogonal views, with cross cuts taken as appropriate.

Left ventricular (LV) volumes, function and mass were measured using a semiautomated

threshold-based technique (CMRtools, Cardiovascular Imaging Solutions, London, UK).

Maximum left atrial (LA) volumes were assessed by a single experienced operator from the

2-chamber (2ch) and 4-chamber (4ch) views at end-systole. This operator was blinded to all

164 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

other clinical and genetic data. The atrial endocardial border was traced to determine LA area

with exclusion of the pulmonary veins and LA appendage. In addition, the LA length was

measured from the midpoint of the mitral annulus plane to the superior aspect of the left

atrium on both views (4ch and 2ch). CMR features of left-ventricular non-compaction

(LVNC) were noted if criteria for LVNC were met (ratio of trabeculated to non trabeculated

myocardium 2.3:1 in diastole(324)).

Septal and lateral left ventricular wall thickness were measured at the level of the papillary

muscles, mid cavity in end-diastole. An average of at least 2 measurements was taken.

Maximum left ventricular wall thickness was measured between basal to apical short axis

images in end-diastole, with the exclusion of papillary muscles. All volume and mass

measurements were indexed to body surface area and referenced to age and gender based

tables(116).

3.3.5 Clinical Data

Detailed phenotyping was completed on all patients, with data entered into a customized

MySQL database via a ‘DCM Events’ form. The data collected is summarized in Table 3-3.

The baseline date for each patient was the date of the first diagnostic cardiac MRI study in

our institution. All participants were interviewed by a research nurse on recruitment whereby

NYHA class was determined, resting heart rate and blood pressure recorded, and self

reported alcohol consumption, ethnicity and medication history were documented.

165 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-3: Data collected on DCM Events form as baseline phenotyping of cohort

Data collected Notes Cardiac MRI LV =left ventricle, RV= right ventricle LV end diastolic volume Data imported from cardiac MRI report; missing data manually entered LV end systolic volume Data imported from cardiac MRI report; missing data manually entered LV stroke volume Data imported from cardiac MRI report; missing data manually entered LV ejection fraction Data imported from cardiac MRI report; missing data manually entered LV mass Data imported from cardiac MRI report; missing data manually entered RV end diastolic volume Data imported from cardiac MRI report; missing data manually entered RV end systolic volume Data imported from cardiac MRI report; missing data manually entered RV stroke volume Data imported from cardiac MRI report; missing data manually entered RV ejection fraction Data imported from cardiac MRI report; missing data manually entered Maximum LV wall Measured from DICOM files; methods outlined in text thickness Mean septal wall thickness Measured from DICOM files; methods outlined in text Mean lateral wall Measured from DICOM files; methods outlined in text thickness Mitral regurgitation Presence and categorical extent (none, mild, moderate, severe) based on MRI report and direct review of images if mild- moderate or moderate-severe. Mid wall gadolinium Presence (and most affected segment) or absence based on MRI report and direct review of images where necessary (e.g. to determine most affected area). Methods outlined in text. Left atrial volume Measured by operator blinded to all other data; methods outlined in text. Coincident myocardial Presence of subendocardial myocardial infarction. infarction (NB: Established coronary artery disease was an exclusion

166 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Data collected Notes criteria, these myocardial infarctions were detected in patients with ‘normal’ coronary arteries or in whom the degree of coronary artery disease was insufficient to account for the degree of LV dysfunction). Pattern of gadolinium Presence or absence of subepicardial LGE, particularly in the enhancement consistent inferolateral segments(325), based on MRI report and direct with chronic myocarditis review of images with senior imaging investigator in borderline cases. Features consistent with Using Petersen criteria(324) of non-compacted to compacted LV non compaction myocardium ratio >2.3 in diastole from long axis SSFP cines. Clinical factors History of myocarditis Review of clinical records prior to baseline date for Cardiologist documented history of myocarditis. History of alcohol excess Review of clinical records prior to baseline date and self reported patient questionnaire; criteria outlined in text History of chemotherapy Review of clinical records prior to baseline date and patient reported questionnaire for any history of chemotherapy use, irrespective of time interval. History of peri-partum Review of clinical records prior to baseline date for the cardiomyopathy development of heart failure with reduced ejection fraction in the last month of pregnancy or first 5 months post partum(326) History of iron overload Review of clinical records prior to baseline date for any Physician documented history of iron overload or clinical records alluding to iron overload (e.g. history of repeated blood transfusions). Inherited muscle disease Review of clinical records prior to baseline date for any Physician documented medical history of inherited muscle disease including review of family pedigree. Pedigree data Family history of DCM Review of family pedigree recorded by research nurse at recruitment as well as clinical records. Definition outlined in text. Family history of sudden Review of family pedigree recorded by research nurse at cardiac death recruitment as well as clinical records. Definition outlined in text. Family history of Review of family pedigree recorded by research nurse at hypertrophic recruitment as well as clinical records

167 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Data collected Notes cardiomyopathy Family history of Review of family pedigree recorded by research nurse at arrhythmogenic right recruitment as well as clinical records ventricular cardiomyopathy ECG/Arrhythmia data Conduction disease Presence and type of any conduction disease including left bundle branch block, 1st, 2nd, and 3rd degree AV block. Baseline arrhythmia Review of clinical records prior to baseline date for history of atrial fibrillation, sustained and non-sustained ventricular tachycardia. Coronary status Presence of normal Review of clinical records prior to baseline date; criteria outlined coronary arteries in DCM diagnosis text. Diagnosis Date of diagnosis Standard Operating Procedure (SOP) developed to determine this information (Appendix). This information was recorded by a team of research nurses.

Expanded definitions of the clinical data collected are provided below.

3.3.5.1 Family history

A family history of dilated cardiomyopathy (DCM in the proband plus at least one other

family member)(327) or sudden cardiac death (in at least 1 family member up to third degree

relative)(49) was recorded.

3.3.5.2 Quantification of alcohol consumption

Weekly alcohol consumption was recorded in units of alcohol. In the UK, 1 unit of alcohol

equals 10mL or 8g of pure alcohol, which is the amount of alcohol an average adult can

process within one hour(328). Consensus medical advice from 1987 to 2016 in the UK

recommended weekly ‘sensible limits’ for alcohol consumption to be no more than 21 units

168 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

(168g) per week for men and 14 units (112g) per week for women(329).

All participants were interviewed by a research nurse and self-reported current weekly

alcohol consumption was recorded. In addition, hospital and primary care medical records

were reviewed for a history of alcohol excess prior to study recruitment. The binary outcome

of ‘alcohol excess’ was defined as any history of the presence or absence of alcohol

consumption greater than 21units (168g)/week for men and 14units (112g)/week for women.

3.3.5.3 Baseline Arrhythmia Phenotyping

The presence of confirmed sustained ventricular tachycardia (VT), non sustained ventricular

tachycardia (NSVT), or atrial fibrillation (AF) at baseline (prior to the date of the first

diagnostic CMR scan at our institution) was collated on all patients using data from Holter

recordings, electrophysiology studies, or clinical records as available from the referring

hospital or primary care provider. NSVT was defined as 3 or more consecutive ventricular

beats lasting up to 30 seconds in duration at a rate greater than 100 beats per minute(330).

Sustained VT was defined as ventricular tachycardia lasting over 30 seconds in duration at a

rate greater than 100 beats per minute. Any AF was included irrespective of duration.

3.3.6 Statistical Analysis

Continuous data are expressed as mean ± standard deviation or median ± interquartile range

and compared using t-tests or Mann-Whitney tests. Categorical data are expressed as number

and percentages, and compared using Fisher’s exact test. Binary baseline data was coded as

confirmed present or absent/missing.

169 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

For the evaluation of predictors of left ventricular mass, univariable linear regression was

performed to identify variables that were predictive of indexed left ventricular mass (LVMi)

at study recruitment. Any variables with p<0.1 from the univariable analysis were considered

for inclusion in an optimised multivariable model, which was created by backward stepwise

selection until only significant variables were included the model. The pre-specified main

analysis was to assess the significance of the addition of TTNtv to this multivariable model

For the evaluation of alcohol as a phenotypic modifier, univariable linear regression was

performed to identify variables that were predictive of left ventricular ejection fraction

(LVEF) at study recruitment. Any variables with p<0.1 from the univariable analysis were

considered for inclusion in an optimised multivariable model, which was created by

backward stepwise selection until only significant variables were included the model. The

pre-specified main analysis for the evaluation of alcohol as a phenotypic modifier was to

assess the significance of the addition of an interaction term between TTNtv and ‘alcohol

excess’ to the optimised multivariable model predicting LVEF. This would test whether

having TTNtv and alcohol excess together has any additional effect other than the effects of

TTNtv and alcohol separately.

For all analyses, a p value ≤ 0.05 was considered statistically significant. All statistical

analysis was conducted in the R environment (version 3.3.1).

170 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4 Results

3.4.1 Cohort Recruitment

The recruitment of patients to the Biobank and curation of the cohort is outlined in Figure

3-1.

Figure 3-1: Outline of DCM cohort recruitment for phenotype study. CMR=cardiovascular magnetic

resonance, HCM=hypertrophic cardiomyopathy, IHD=ischaemic heart disease, LVH=left ventricular

hypertrophy, NAD=nothing abnormal detected, FS_NAD=normal family screen, ARVC=arrhythmogenic

right ventricular cardiomyopathy, Aorta=primary aortic pathology, LVNC=left ventricular non-

compaction.

As outlined, in the first step, 928 patients had an imaging (echo or CMR) diagnosis of DCM.

The phenotype analysis was restricted to patients with CMR only, to enable the quantitative

171 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

in depth phenotyping that CMR affords. The final cohort for phenotype analysis comprised

732 patients with DCM confirmed by CMR, after exclusion of 39 patients with coronary

artery disease meeting exclusion criteria, 19 patients who had not had coronary artery

assessment but who were found to have myocardial infarction on CMR, and 8 patients in

whom bystander myocardial infarction status was unknown (for example, if gadolinium

contrast could not be administered).

3.4.2 Patient demographics

The final phenotype study population consisted of 732 patients with a clinical and imaging

diagnosis of DCM confirmed on CMR. There were 475 men (64.9%) and the majority of the

cohort was Caucasian (n=618, 84.4 %). The mean age of the cohort at recruitment was 53.4

years (±14.4; range 11.5 to 88.4 years). The majority of patients were in NYHA class I

(n=306, 41.8 %) or II (n=285, 38.9 %) at recruitment. A family history of DCM was found in

117 patients (16.0 %) and a family history of sudden cardiac death was found in 109 patients

(14.9 %). Of these, 41 patients (5.6 %) had both a family history of DCM and sudden cardiac

death. A summary of demographics, clinical factors and baseline medication are shown in

Table 3-4.

172 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-4: Baseline demographics in DCM cohort. Age is summarized as mean ± standard deviation. The

remaining data are categorical and summarised as count (percentages).

Variable Total cohort, n=732 Age at baseline scan (years) (mean (sd)) 53.4 (14.40) Race (self-reported) Afro-Caribbean 27 (3.7) African 20 (2.7) Asian 39 (5.3) Caucasian 618 (84.4) Chinese 3 (0.4) Mixed 4 (0.5) Other 21 (2.9) Male gender 475 (64.9) NYHA class 1 306 (43.7) 2 285 (40.7) 3 101 (14.4) 4 8 (1.1) Family history of DCM 117 (16.0) Family history of sudden cardiac death 109 (14.9) Clinical history of myocarditis 29 (4.0) Alcohol excess 114 (15.6) Chemotherapy 35 (4.8) Inherited muscle disease 4 (0.5) Peripartum cardiomyopathy 9 (1.2) Beta blocker use 516 (70.5) ACE inhibitor use 582 (79.5) Aldosterone Antagonist use 264 (36.1) Diuretic use 333 (45.5)

173 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.2.1 Time interval since DCM diagnosis

For the majority of patients (n=381, 52.0%), the CMR scan at recruitment was also the first

confirmation of diagnosis. For the remaining patients, the interval between date of diagnosis

and the baseline study date is shown in Figure 3-2. Of these patients, the majority

(n=270/351, 76.9%) were diagnosed with DCM within the 2 years preceding their study

baseline CMR date.

Figure 3-2: Interval between date of diagnosis and baseline date for patients for whom the CMR study at

recruitment was not the diagnostic study. Right hand panel shows the interval limited to 2 years.

3.4.3 Overall DCM phenotype

Ventricular volumes and function for the cohort are illustrated in Figure 3-3 and the plots for

left ventricular ejection fraction and indexed volumes for patients with DCM compared to

healthy controls are shown in Figure 3-4. The baseline cardiac MRI phenotype is summarised

in Table 3-5.

174 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Overall, this was a cohort with moderately impaired left ventricular function, consistent with

the entry criteria of the diagnostic CMR scan. Mean left ventricular ejection fraction

(LVEF,%) was 39.0 ± 12.45 (median=40%, interquartile range 30 to 49%). There is an

approximately even distribution of patients with LVEF above and below 40%. One third of

the cohort had severely impaired left ventricular function with LVEF <35% (n=267, 36%).

Mean left ventricular end diastolic volume indexed to body surface area (LVEDVi) was

127.3± 35.9 mL/m2 (interquartile range 102.9 to 143.7mL/m2) and mean left ventricular end

systolic volume indexed to body surface area (LVESVi) was 80.2 ± 36.1 mL/m2 (interquartile

range 53.7 to 97.8 mL/m2). Left and right ventricular end diastolic and end systolic volumes

are positively skewed, reflecting a small proportion of individuals with very dilated

chambers. This is also reflected in the distribution of left ventricular mass, as elevated left

ventricular mass reflects increased ventricular dilation associated with eccentric hypertrophy.

Stroke volume is calculated as the difference between end diastolic and end systolic volume,

therefore this has a normal distribution in our cohort as both the end diastolic and end systolic

volumes are skewed in the same direction and are correlated within each patient.

Mean right ventricular ejection fraction (RVEF, %) was 51.2 ± 13.9 % (median= 53.0 %,

interquartile range 43.0 to 61.0%). Approximately one quarter of the cohort had evidence of

right ventricular systolic dysfunction, with RVEF <45% (n=194, 27%).

175 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-3: Histograms showing the distribution of left and right ventricular function and volumes,

indexed to body surface area. L/RVEF=left/right ventricular ejection fraction, L/RVEDVi=indexed

left/right ventricular end diastolic volume, L/RVESVi= indexed left/right ventricular end systolic volume,

L/RVSVi=indexed left/right ventricular stroke volume, LVMi=indexed left ventricular mass.

176 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-4: CMR phenotype of DCM cohort compared to controls. Three plots show values for LVEF

(left ventricular ejection fraction), LVEDVi (indexed left ventricular end diastolic volume) and LVESVi

(indexed end systolic volume) for patients with DCM (red dots) compared to controls (blue dots). The

control participants, described in the methods, are included for illustrative purposes only and are taken

from a previously published cohort from our institution(116).

Mid-wall myocardial fibrosis, detected through LGE-CMR, was found in 256 patients (35%)

and the vast majority was located in the septum (n=188, 73%) (Table 3-5). This was a cohort

of patients with no known coronary artery disease yet a small proportion of patients had

bystander myocardial infarction (n=17, 2.3%). Overall, of the 336 patients with mitral

regurgitation, the majority of patients had mild regurgitation only (n=239, 33% of overall

cohort) (Table 3-5).

177 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-5: Baseline CMR phenotype in DCM cohort. Continuous data are summarized as mean ±

standard deviation. Categorical data are summarized as count and percentages.

CMR Phenotype DCM, n=732 Left ventricular ejection fraction (%) 39.0 (12.5) Indexed LV end diastolic volume (mL/m2) 127.3 (35.9) Indexed LV end systolic volume (mL/m2) 80.2 (36.1) Indexed LV stroke volume (mL/m2) 47.1 (13.3) Indexed LV mass (g/m2) 90.5 (26.0) Right ventricular ejection fraction (%) 51.7 (13.9) Indexed RV end diastolic volume (mL/m2) 87.9 (24.3) Indexed RV end systolic volume (mL/m2) 44.1 (21.9) Indexed RV stroke volume (mL/m2) 43.9 (12.7) CMR features of LV non-compaction 36 (4.9) Mid wall late gadolinium enhancement 256 (35.0) Mid wall late gadolinium enhancement most affected segment Anterior 8 (3.1) Inferior 18 (7.0) Lateral 44 (17.1) Septal 188 (72.9) Late gadolinium enhancement consistent with myocarditis No 671 (91.7) Unknown 12 (1.6) Yes 49 (6.7) Mitral regurgitation Mild 239 (32.7) Moderate 81 (11.1) None 396 (54.1) Severe 16 (2.2) Indexed left atrial volume (mL/m2) 60.3 (25.4) Maximum LV wall thickness (mm) 9.8 (2.2) Mean septal wall thickness (mm) 7.8 (1.9) Mean lateral wall thickness (mm) 5.5 (1.6)

178 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Mean indexed left atrial volume (LAVi) was 60.2 ± 25.4 mL/m2. The mean maximum left

ventricular wall thickness was 9.8mm and mean septal and lateral wall thickness at mid

cavity level were 7.8mm and 5.5mm respectively. The correlation between these markers and

other indices of cardiac structure and function are shown in Figure 3-5.

Figure 3-5: CMR phenotype correlation plot. Figure shows scatter plots, Pearson’s correlation

coefficients and significance thresholds for correlation between major indices of cardiac structure and

function. P-values: ***<0.001, **<0.01, *<0.05, <0.1, blank for p≥0.1.

This shows that LVEF is highly negatively correlated with LVESVi and LVEDVi, and

positively correlated with LVSVi, which is expected and demonstrates that left ventricular

dilation is associated with impaired left ventricular function in this cohort. As also expected,

LVEF is also highly positively correlated with RVEF and RVSVi, but not RVEDVi. It is

however interesting to note that the correlation between LVEF and RVEF is not perfectly

179 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

linear (r=0.59) and that patients with impaired left ventricular function may still have

relatively preserved right ventricular function.

LAVi is weakly correlated to LVEF, RVEF, LVEDVi, LVESVi, RVEDVi, RVESVi, and

LVMi, suggesting that the biological basis for left atrial dilation may be distinct from

ventricular pathology, and highlights the utility of LAVi as an independent marker of cardiac

structure and function.

Finally, maximal left ventricular wall thickness is as expected positively correlated with left

ventricular mass. However, it is not correlated with LVEF or LVEDVi, suggesting that wall

thickness and the eccentric cardiac hypertrophy of DCM in this cohort (an increase in left

ventricular mass not accompanied by an increase in wall thickness) are not governed purely

by the extent of cardiac dilatation. I explore this further in the analysis of titin

cardiomyopathy.

3.4.4 Curation of titin truncating variants

Building upon the results from Chapter 2, I curated the truncating variants in titin (TTNtv) in

this cohort. The process is outlined in Figure 3-6.

180 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-6: Flowchart outlining filtering steps to curate TTNtv for phenotype analysis. Details outlined in

text. TTNtv= truncating variants in the titin gene, ExAC=Exome Aggregation Consortium dataset,

QD=QualbyDepth score, PAV=Protein altering variants, tv=truncating variants, ms=missense variants,

PSI=percentage spliced in, IGV=Integrative Genomics Viewer, CMR=cardiovascular magnetic

resonance.

181 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Initially, all patients with an imaging diagnosis of DCM (n=928) underwent targeted

sequencing on a customised cardiac gene panel as outlined in Chapter 2. This identified

65469 protein altering variants in 888 patients. The DNA from 28 patients could not be

sequenced due to insufficient sample or poor sample quality. In these patients, we did not

have blood, but only saliva samples.

The majority of the remaining samples were sequenced on the TruSight Cardio Assay and

NextSeq platform (n=826, 89%) (Table 3-6).

Table 3-6: Sequencing platform and assay used in DCM cohort. ICC_169 was the precursor assay to

TruSight Cardio.

Assay and Platform Number of patients Number of patients Confirmed sequenced in overall sequenced in CMR TTNtv DCM cohort cohort N=928 N=732 TruSightCardio and NextSeq 826 660 102 TruSightCardio and MiSeq 47 34 6 ICCv6 and HiSeq 5 5 0 ICCv5 and 5500xl 4 3 0 ICCv4_TopUp and 5500xl 13 12 0 ICC_169 and MiSeq 5 2 0 Not sequenced 28 16 -

For patients with TTNtv, the percentage coverage at 30x read depth per case is shown in

Figure 3-7. All samples had >95% of reads with at least 30x read depth.

182 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-7: Percentage coverage of TTN at 30x coloured by assay for patients with TTNtv. ICC_169 was

the precursor assay to TruSight Cardio (TS Cardio). All samples were sequenced on Illumina platforms.

From the 65,469 protein altering variants in 888 patients, filtering based on the target gene

list outlined in Table 3-1 resulted in 25,468 protein altering variants in 888 people. Filtering

for rare variants only, defined as a minor allele frequency <0.0001 in the ExAC dataset,

resulted in 787 protein altering variants in 503 people. Filtering on sequencing metrics

(minimum coverage of 10x, QD score over 4, and allelic balance over 20%), left 738 protein

altering rare variants in 491 people.

Then filtering by variant class (Table 3-1), left 265 disease specific variants in 240 people.

This included 117 TTNtv which were curated further, retaining only variants in constitutively

expressed exons (PSI >90%, n=112 TTNtv in 109 patients).

The TTNtv were then confirmed with Sanger sequencing or interactive review of mapped

reads using Integrative Genomics Viewer (IGV). During this process, 2 TTNtv variants were

excluded in 1 patient. These variants had each been annotated by both Unified Genotyper and

Haplotype Caller as frameshift truncating variants.

183 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

c.74761_74767delTCCAGGG

c.74759_74760insA

However, these variants are a 7bp deletion and a 1 base pair insertion in an adjacent region

resulting in an inframe deletion and an amino acid substitution. The IGV output is shown in

Figure 3-8. Whilst these are clearly protein altering variants, they do not meet the criteria for

truncating variants and therefore they were excluded from the primary titin phenotype cohort.

Figure 3-8: IGV output of individual with 2 protein altering variants in TTN, one 7 base pair deletion and

1 base pair insertion.

There were 2 further patients (related) who carried only one TTNtv which had been

incorrectly called as two variants by the variant callers. These individuals were reclassified as

having one TTNtv each:

(1) c.100943_100944delGA Chr. 2: 179400398-179400399 (reverse strand). Called as

frameshift mutation.

(2) c.100942A>T Chr. 2: 179400400 (reverse strand). Called as a nonsense mutation.

184 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

At this position (chr2: 179400398-179400400), the reference genome is TCT. The effect of

the 2 variants shown is TC-> and T>A. Therefore at this 3 base pair position, the overall

change is TCT->A, which is a 2bp deletion and a frameshift.

This left 108 confirmed TTNtv in 108 patients across the entire DCM cohort (n=928 with an

imaging diagnosis of DCM). Of these, 57 were confirmed through review on IGV and 51

were confirmed by Sanger sequencing.

In the subset of 716 patients with DCM confirmed by CMR who had undergone sequencing,

there were 83 confirmed rare TTNtv in constitutively expressed exons in 83 patients. The

variants are listed in Table 3-7. They are plotted according to protein domain in Figure 3-9.

Genotype-phenotype analysis was then performed in this cohort.

185 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-7: List of truncating variants in TTN gene (n=83). All variants annotated to titin meta-transcript ENST00000589042. Callers= U, Unified genotyper or

H=Haplotype caller. PSI= percentage spliced in, a measure of splicing derived from RNAseq, is an estimate of the percentage of TTN transcripts that incorporate a

given exon.

cDNA Variant Variant type Callers Coverage Allelic Confirmed ExAC ExAC TTN TTN band TTN PSI

Balance frequency count exon

c.45852_45853delCA frameshift U,H 984 0.5 IGV 0 0 248 I-band 100

c.79539T>A nonsense U,H 999 0.53 IGV 0 0 327 A-band 100

c.55037delG frameshift U,H 819 0.51 IGV 0 0 284 A-band 100

c.83635delA frameshift U,H 998 0.49 IGV 0 0 327 A-band 100

c.45812T>G nonsense U,H 999 0.43 Sanger 0 0 248 I-band 100

c.82513delA frameshift U,H 999 0.51 Sanger 0 0 327 A-band 100

c.78507delT frameshift U,H 1000 0.46 Sanger 0 0 327 A-band 100

c.6790+1G>T essential U,H 1000 0.55 Sanger 0 0 29 I-band 100

splice site

c.81262_81269delCAGATGCT frameshift U,H 927 0.46 Sanger 0 0 327 A-band 100

c.97062delA frameshift U,H 604 0.43 IGV 0 0 349 A-band 99

c.12757C>T nonsense U,H 1000 0.51 Sanger 0 0 49 I-band 99

c.55525_55531delGACAGGA frameshift U,H 999 0.46 Sanger 0 0 288 A-band 100

186 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

cDNA Variant Variant type Callers Coverage Allelic Confirmed ExAC ExAC TTN TTN band TTN PSI

Balance frequency count exon

c.43792delG frameshift U,H 1000 0.5 Sanger 0 0 238 I-band 100

c.78184G>T nonsense U,H 1000 0.48 Sanger 0 0 327 A-band 100

c.10303+2T>C essential U,H 618 0.52 IGV 8.25E-06 1 44 I-band 100

splice site

c.59226T>G nonsense U,H 967 0.53 IGV 0 0 301 A-band 100

c.101996G>A nonsense U,H 644 0.52 Sanger 0 0 359 M-band 100

c.60931C>T nonsense U,H 999 0.52 Sanger 0 0 305 A-band 100

c.98506C>T nonsense U,H 1000 0.47 Sanger 0 0 353 A-band 100

c.48527G>A nonsense U,H 999 0.46 Sanger 0 0 260 A-band 100

c.45307C>T nonsense U,H 1000 0.47 IGV 0 0 246 I-band 100

c.100644dupT frameshift U,H 988 0.52 IGV 0 0 358 A-band 99

c.50170C>T nonsense U,H 502 0.54 Sanger 0 0 267 A-band 100

c.69630C>A nonsense U,H 999 0.51 Sanger 0 0 326 A-band 100

c.52035_52036insTT frameshift U,H 1000 0.57 Sanger 0 0 274 A-band 100

c.84120delT frameshift U,H 999 0.5 IGV 0 0 327 A-band 100

c.64449dupA frameshift U,H 1000 0.52 IGV 0 0 310 A-band 100

c.58732+2T>C essential U,H 1000 0.47 Sanger 0 0 299 A-band 100

187 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

cDNA Variant Variant type Callers Coverage Allelic Confirmed ExAC ExAC TTN TTN band TTN PSI

Balance frequency count exon

splice site

c.55525_55531delGACAGGA frameshift U,H 999 0.48 Sanger 0 0 288 A-band 100

c.8307_8308delTG frameshift U,H 998 0.45 Sanger 0 0 35 I-band 100

c.63025C>T nonsense U,H 999 0.51 Sanger 0 0 305 A-band 100

c.81321C>G nonsense U,H 999 0.47 Sanger 0 0 327 A-band 100

c.50170C>T nonsense U,H 636 0.46 Sanger 0 0 267 A-band 100

c.47875+1G>A essential U,H 999 0.51 Sanger 0 0 256 A-band 100

splice site

c.47697C>A nonsense U,H 1000 0.51 Sanger 0 0 255 A-band 100

c.86967G>A nonsense U,H 999 0.49 Sanger 0 0 328 A-band 100

c.67567delG frameshift U,H 999 0.47 IGV 0 0 320 A-band 100

c.4724_4728delTGAAA frameshift U,H 998 0.48 Sanger 0 0 27 near Z-disk 100

c.12643_12644delCA frameshift U,H 839 0.52 Sanger 0 0 49 I-band 99

c.92683C>T nonsense U,H 528 0.45 Sanger 0 0 340 A-band 100

c.86641delC frameshift U,H 1000 0.46 Sanger 0 0 327 A-band 100

c.93291_93301delTGTTGGTGA frameshift U,H 1000 0.52 Sanger 0 0 340 A-band 100

GC

188 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

cDNA Variant Variant type Callers Coverage Allelic Confirmed ExAC ExAC TTN TTN band TTN PSI

Balance frequency count exon

c.52223_52227dupAGAAA frameshift U,H 1000 0.53 Sanger 0 0 275 A-band 100

c.76666_76684dupATAATTGAT frameshift U,H 999 0.45 Sanger 0 0 327 A-band 100

GTCACTAGCA

c.100943_100944delGA frameshift U,H 996 0.47 IGV 0 0 359 M-band 100

c.51781C>T nonsense U,H 1000 0.52 IGV 0 0 274 A-band 100

c.61855delG frameshift U,H 1000 0.52 Sanger 0 0 305 A-band 100

c.6322G>T nonsense U,H 1000 0.46 IGV 0 0 28 near Z-disk / 100

I-band

c.51965dupT frameshift U,H 828 0.5 IGV 0 0 274 A-band 100

c.100267_100268delAA frameshift U,H 994 0.49 IGV 0 0 358 A-band 99

c.83928dupT frameshift U,H 1000 0.52 Sanger 0 0 327 A-band 100

c.3380+1G>C essential U,H 341 0.41 IGV 0 0 20 near Z-disk 100

splice site

c.47697C>A nonsense U,H 721 0.53 IGV 0 0 255 A-band 100

c.86821+2T>A essential U,H 717 0.49 IGV 8.91E-06 1 327 A-band 100

splice site

c.41473C>T nonsense U,H 955 0.49 Sanger 0 0 227 I-band 100

189 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

cDNA Variant Variant type Callers Coverage Allelic Confirmed ExAC ExAC TTN TTN band TTN PSI

Balance frequency count exon

c.53881+1G>T essential U,H 985 0.47 Sanger 0 0 280 A-band 100

splice site

c.106629delA frameshift U,H 999 0.5 Sanger 0 0 361 M-band 100

c.87716delG frameshift U,H 366 0.56 Sanger 0 0 330 A-band 100

c.82240C>T nonsense U,H 999 0.51 IGV 1.66E-05 2 327 A-band 100

c.53206C>T nonsense U,H 1000 0.45 IGV 0 0 278 A-band 100

c.89216delC frameshift U,H 995 0.54 IGV 0 0 335 A-band 100

c.41447delG frameshift U,H 674 0.5 Sanger 0 0 227 I-band 100

c.89750dupG frameshift U,H 999 0.49 Sanger 0 0 336 A-band 100

c.85090C>T nonsense U,H 999 0.52 Sanger 0 0 327 A-band 100

c.102958delA frameshift U,H 1000 0.54 IGV 0 0 359 M-band 100

c.43602_43615delGCGCCTACA frameshift U,H 803 0.55 Sanger 0 0 237 I-band 100

CACCA

c.100943_100944delGA frameshift U,H 1000 0.41 Sanger 0 0 359 M-band 100

c.75250C>T nonsense U,H 1000 0.47 IGV 0 0 327 A-band 100

c.94721_94722delTC frameshift U,H 997 0.49 IGV 0 0 342 A-band 100

c.63025C>T nonsense U,H 997 0.49 IGV 0 0 305 A-band 100

190 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

cDNA Variant Variant type Callers Coverage Allelic Confirmed ExAC ExAC TTN TTN band TTN PSI

Balance frequency count exon

c.94978dupG frameshift U,H 999 0.47 IGV 0 0 343 A-band 100

c.41473C>T nonsense U,H 473 0.54 IGV 0 0 227 I-band 100

c.40791dupG frameshift U,H 468 0.49 Sanger 0 0 224 I-band 100

c.63025C>T nonsense U,H 998 0.51 IGV 0 0 305 A-band 100

c.12010G>T nonsense U,H 584 0.48 IGV 0 0 49 I-band 99

c.83416C>T nonsense U,H 999 0.5 IGV 0 0 327 A-band 100

c.59201_59202delCT frameshift U,H 379 0.55 IGV 8.28E-06 1 301 A-band 100

c.73846C>T nonsense U,H 680 0.53 IGV 0 0 327 A-band 100

c.83416C>T nonsense U,H 658 0.48 IGV 0 0 327 A-band 100

c.58870C>T nonsense U,H 657 0.48 IGV 0 0 300 A-band 100

c.52254G>A nonsense U,H 377 0.47 IGV 0 0 275 A-band 100

c.44364delC frameshift U,H 250 0.52 IGV 0 0 241 I-band 100

c.55525_55531delGACAGGA frameshift S,U 250 0.23 Sanger 0 0 288 A-band 100

191 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

| | || | | || || ||||| ||| | ||||| || || || | | | | | | | ||| |||||||| |||| || || || | | ||||| | n=83 DCM, domains TTN protein PSI 0

0 35991

Figure 3-9: Position of TTNtv variants according to protein domains and PSI (percentage spliced in). The colours of the protein domain cartoon correspond to: Z

disc =red, I band=blue, A band=green, M-line=purple. Only variants in constitutive exons are included in analysis. In line with previous data, the majority of

variants in DCM patients lie in the distal I band and A band.

192 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.5 TTNtv cardiomyopathy phenotype

3.4.5.1 Demographics

Across the cohort of patients with DCM confirmed by CMR who underwent DNA

sequencing, 11.6% of patients had TTNtv (n=83/716). Patients with TTNtv were younger at

study recruitment compared to patients without TTNtv (49.0 vs 54.1 years, p =0.002). There

was no significant difference in gender or ethnicity between groups (Table 3-8).

As expected for a genetic condition, patients with TTNtv were more likely to have a family

history of DCM (31% vs 14%, p<0.001; Table 3-8). Amongst the TTNtv patients with a

family history of DCM (n=26), 22 in this cohort were probands and 4 were affected family

members. For the purposes of phenotype analysis, the 4 affected family members were

retained in the cohort. In two of the three families, the affected relatives had a milder

phenotype than the proband, though the sample size prohibited further statistical analysis

(Figure 3-10). The affected family members of the remaining 22 probands with a family

history of DCM had not been recruited to the study.

Figure 3-10: Cardiac phenotype in probands and affected relatives. Cardiac phenotype of proband (x axis

Proband Status =1) and affected relatives (x axis Proband Status =0) in 3 families with DCM. Colour of

points refers to one family. LV/RV=left/right ventricle; EDVi, indexed end-diastolic volume (mL/m2); EF,

ejection fraction (%).

193 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

There was no difference in family history of sudden cardiac death, suggesting that TTNtv

may not be associated with adverse arrhythmic outcomes (Table 3-8). There was significantly

less left bundle branch block (LBBB) and less evidence of conduction disease in patients with

TTNtv compared to patients without TTNtv (Table 3-8).

Table 3-8: Baseline demographics in DCM patients stratified by TTNtv status. Age is summarised as

mean (standard deviation) and compared with t-test. The remaining categorical variables are

summarised as count and percentages and compared using Fisher’s exact test. * indicates significance

after correction for multiple testing (p<0.004).

TTNtv Absent TTNtv Present P value N=633 N=83 Age at baseline date (years) 54.1 (14.33) 49.0 (13.45) 0.002 Male gender 410 (64.8) 59 (71.1) 0.27 Race Caucasian 533 (84.2) 75 (90.4) 0.39 Family history of DCM 87 (13.7) 26 (31.3) <0.001* Family history of sudden cardiac death 95 (15.0) 12 (14.5) 1.0 Conduction disease 217 (34.3) 18 (21.7) 0.03 Left bundle branch block 179 (28.5) 8 (9.8) <0.001* History of myocarditis 27 (4.3) 1 (1.2) 0.24 History of peripartum cardiomyopathy 6 (0.9) 1 (1.2) 0.58 History of chemotherapy 32 (5.1) 2 (2.4) 0.41 History of alcohol excess 98 (15.5) 13 (15.7) 1.0 NYHA class 0.17 1 261 (43.1) 39 (49.4) 2 245 (40.4) 32 (40.5) 3 94 (15.5) 6 (7.6) 4 6 (1.0) 2 (2.5) Unknown 27 (4.3) 4 (4.8)

194 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.5.2 CMR phenotype

Upon evaluation of cardiac structure and function, there was no significant difference in

biventricular ejection fraction in patients with and without TTNtv (LVEF (%; median [IQR])

TTN+/-; 39.0 [27.0, 49.0] vs 40.0 [30.0, 49.0], p=0.48, RVEF (%; median [IQR]) TTN+/-;

53.0 [44.0, 62.0] vs 54.0 [44.0, 61.0], p=0.65) (Table 3-9). After correction for multiple

testing, there were no significant differences in biventricular stroke volume or cardiac

chamber dimensions (left and right ventricular volumes or left atrial volume) between

patients with TTNtv and patients without TTNtv (Table 3-9).

There was no detectable difference in the frequency of CMR features of left ventricular non-

compaction between patients with TTNtv. Whilst the numbers of affected cases are too small

for definitive analysis, in this large cohort of patients with TTNtv, this suggests that the

LVNC phenotype in DCM is not driven by TTNtv.

There was no significant difference in the pattern of gadolinium enhancement seen in patients

with TTNtv compared to DCM patients without TTNtv (Table 3-9). Specifically, there was

no difference in the presence or location of left ventricular mid wall fibrosis, an important

prognostic marker detected in one third of cases with DCM. There was no difference in the

presence of bystander myocardial infarction or gadolinium enhancement consistent with

chronic myocarditis between TTNtv positive and negative groups, though there were limited

cases with these features limiting the scope of this analysis.

195 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-9: Baseline CMR variables in DCM cohort stratified by TTNtv status. Continuous data are summarised as mean (standard deviation) or median [interquartile range] and compared using Mann- Whitney tests. Categorical variables are summarised as count and percentages and between group differences tested with Fisher’s exact test. *significant after adjusting for multiple testing (p≤0.003)

TTNtv Absent TTNtv Present P value N=633 N=83 Left ventricular ejection fraction (%) 40.0 [30.0, 49.0] 39.0 [27.0, 49.0] 0.48

Indexed LV end diastolic volume (mL/m2) 118.5 [103.5, 143.8] 111.1 [100.6, 136.4] 0.14

Indexed LV end systolic volume (mL/m2) 68.9 [53.8, 96.7] 67.37 [51.0, 100.1] 0.52

Indexed LV stroke volume (mL/m2) 47.8 (13.3) 43.6 (12.8) 0.007 Indexed LV mass (g/m2) 87.4 [73.6, 106.5] 81.3 [69.4, 96.8] 0.02 Right ventricular ejection fraction (%) 54.00 [44.0, 61.0] 53.0 [44.0, 62.0] 0.65

Indexed RV end diastolic volume (mL/m2) 84.7 [70.5, 101.8] 84.9 [70.1, 99.4] 0.61

Indexed RV end systolic volume (mL/m2) 40.0 [28.8, 53.0] 41.0 [28.0, 51.5] 0.96

Indexed RV stroke volume (mL/m2) 44.2 (12.7) 41.7 (13.1) 0.10 Indexed left atrial volume (mL/m2) 54.1 [43.9, 68.4] 57.17 [42.4, 70.6] 0.79

Maximum LV wall thickness (mm) 9.9 (2.2) 8.8 (1.8) <0.001* Mean septal wall thickness (mm) 7.9 (1.9) 7.2 (1.5) 0.002* Mean lateral wall thickness (mm) 5.6 (1.6) 5.0 (1.2) 0.001* CMR features of LVNC 31 (4.9) 5 (6.0) 0.60 Mid wall late gadolinium enhancement 224 (35.4) 26 (31.3) 0.54 present Mid wall late gadolinium enhancement 0.32 most affected segment

Anterior 8 (3.6) 0 (0.0) Inferior 17 (7.6) 1 (3.7) Lateral 42 (18.7) 2 ( 7.4) Septal 158 (70.2) 24 (88.9) Late gadolinium enhancement consistent 46 (7.3) 2 (2.4) 0.08 with myocarditis

Mitral regurgitation 0.48 None 340 (53.7) 49 (59.0) Mild 210 (33.2) 23 (27.7) Moderate 66 (10.4) 11 (13.3) Severe 15 (2.4) 0 (0.0)

196 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Notably however, patients with TTNtv had thinner left ventricular walls and a trend towards

lower indexed left ventricular mass (LVMi) in the absence of statistically significant

differences in left ventricular dilation, compared to patients without TTNtv (Figure 3-11,

Table 3-9).

Figure 3-11: TTNtv hypertrophy phenotype. Beeswarm and boxplot (black bars showing median) of

indexed left ventricular (LV) mass, mean septal wall thickness and indexed left ventricular end diastolic

volume stratified by titin status, showing that patients with TTNtv have lower indexed LV mass and

thinner LV walls in the absence of evidence of differences in LV dilatation. Between group comparisons

are made using the Mann Whitney test.

This suggests that TTNtv cardiomyopathy is associated with a different hypertrophic

response compared to patients without TTNtv. To explore this further, linear regression

analysis was performed, evaluating predictors of left ventricular mass.

3.4.5.3 TTNtv and LV mass

3.4.5.3.1 Baseline predictors of left ventricular mass

On univariable linear regression analysis, potential predictors of LVMi were evaluated. To

build the baseline model, before evaluating the importance of TTNtv, biological and clinical

197 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

variables that could plausibly predict LV mass were selected. These were age, gender, a

history of hypertension, race, systolic blood pressure (as measured on the day of recruitment),

and CMR features of LVNC (could contribute to increased LV mass). LVEDVi or LVESVi

were not included as they are strongly correlated with LVMi (LVEDVi, r=0.58, p<0.00001;

LVESVi, r=0.57, p<0.00001).

Initially, univariable linear regression was performed on each of these predictors against

LVMi and variables that were significant at a p value threshold of <0.10 were retained. Of

these variables, gender, race, and a history of hypertension were significant (Table 3-10).

Table 3-10: Variables evaluated on univariable linear regression as predictors of left ventricular mass

LVMi= indexed left ventricular mass.

Variable Estimate change P value Lower CI Upper CI in LVMi (g/m2) Age (per 1 year) 0.1 0.25 -0.1 0.2 Male Gender 17.7 <0.0001 13.9 21.5 Race (compared to Afro-Caribbean) African 6.0 0.46 -10.0 22.1 Asian -15.3 0.02 -28.2 -2.3 Caucasian -9.7 0.06 -19.8 0.5 Chinese -6.1 0.70 -36.9 24.7 Mixed -0.8 0.96 -27.9 26.4 Other -8.8 0.26 -24.1 6.4 History of hypertension 6.8 0.001 2.7 11.0 Systolic blood pressure (per 1mmHg) 0.0 0.76 -0.1 0.1 CMR features of LVNC -3.7 0.42 -12.6 5.3

198 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

The full model was built including these 4 variables and then optimised using reverse

stepwise selection, with a p value threshold for exclusion of <0.05. At this stage race was

removed, leaving gender and hypertension. Each of the previously discarded variables was

then added in turn, both by themselves (systolic blood pressure, age, LVNC) and together

with race to assess if any variable either became significant in the context of other variables,

or whether their inclusion affected the coefficient or significance of the existing variables.

Race was also evaluated collapsed into Caucasian ethnicity or not. These discarded variables

remained non significant and did not have an effect on the retained variables of gender and

hypertension.

3.4.5.3.2 The effect of TTNtv on left ventricular mass

The initial plots (Figure 3-11) suggested that TTNtv might be associated with a blunted

hypertrophic response. This means that TTNtv are associated with a reduced LVMi and

thinner LV walls, in the absence of evidence of significant differences in LV dilation. To

explore the relationship between LVMi and LVEDVi further, the regression of

LVMi~LVEDVi, stratified by TTNtv status was plotted (Figure 3-12). This shows that for

any given increase in LVEDVi, patients with TTNtv had a reduced increased in LVMi

(Increase in LVMi per 1mL/m2 increase in LVEDVi: 0.26mg/m2 for patients with TTNtv,

0.42mg/m2 for patients without TTNtv).

199 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-12: LV mass and LV end diastolic volume. Indexed LV mass (LVMi) plotted against indexed

end diastolic LV volume (LVEDVi). Red points are patients with TTNtv. The regression slope of LVMi

~LVEDVi (LVMi increase per 1mL/m2 increase in LVEDVi) is shown for patients with TTNtv (red line:

0.26g/m2) and without TTNtv (black line: 0.42g/m2). Typically, LVMi increases in line with increasing

indexed left ventricular end diastolic volume but this effect is reduced in TTNtv DCM (p=0.006).

On univariable linear regression analysis, the presence of TTNtv was associated with a

7.8g/m2 reduction in LVMi (95% CI -13.7 to -1.8g/m2, p=0.01). On multivariable analysis,

TTNtv remained predictive of a lower LVMi after adjusting for the covariates in the baseline

model, with an estimated 7.8g/m2 reduction in LVMi (adjusted p=0.007) (Table 3-11 and

Figure 3-13).

Table 3-11: Multivariable regression analysis evaluating predictors of left ventricular mass

Variable Estimate change in LVMi (g/m2) P value Lower CI Upper CI

Male gender 18.0 <0.0001 14.2 21.8

History of hypertension 6.2 0.002 2.2 10.2

Presence of TTNtv -7.8 0.007 -13.4 -2.1

200 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-13: Forest plot showing the predictors of left ventricular mass, demonstrating that TTNtv are

associated with a lower LV mass after adjustment for gender and hypertension.

3.4.5.3.3 Model checking

The adjusted R2 of the final model was 0.12. This is low and suggests that the model explains

little of the overall variation in LVMi, which is not unexpected given that the model includes

only 3 binary predictors. However, all potential clinical variables were evaluated and

considered for inclusion in the model. In this cohort, there are as yet unidentified predictors

of LVMi.

Allowing for this, the model as it stands does demonstrate that TTNtv have an influence on

LVMi, which is a potentially important finding. Therefore I checked the model to evaluate

whether assumptions of linearity and homoscedasticity have been met and potential

influential outliers identified (Figure 3-14). These plots show that there is no clear pattern in

the distribution of the residuals and that assumptions of linearity and equal variance are valid.

However, whilst there are no points that are clear outliers, data row 392 is flagged for further

review.

201 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-14: Model checking of final linear regression model predicting LV mass (LVMi~gender,

hypertension, + TTNtv). The Residuals vs Fitted plot shows that the residuals are equally spread above

and below the horizontal red line, without distinct patterns. This suggests that there are no underlying

non-linear relationships that have not been accounted for. The QQ plot shows whether the residuals are

normally distributed, and here the residuals appear to fit the expected distribution, with the exception of

data row 392. In the Scale-Location plot, the residuals appear to be spread equally across a range of

predictors, thereby meeting the assumption of equal variance (homoscedasticity). In the final Residuals vs

Leverage plot, checking for influential observations, there are no points outside of Cook’s distance

(dashed red line not seen on plot). However, data row 392 is again highlighted, suggesting it may be

influential to the regression results.

On review of the data, this individual is a 69 year old female, with no history of hypertension,

with a LVMi of 82.5g/m2, which is within a clinically appropriate range, and no TTNtv. No

inaccuracies in data collection and recording were identified. The model was re-run after

excluding this individual and there was little change to the regression coefficient for the

effect of TTNtv on LVMi (TTNtv present: -7.8g/m2 change in LVMi, 95% CI -13.4 to -2.1,

p=0.008). The original final model therefore appears reasonable.

202 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.5.3.4 Sensitivity analysis: other sarcomeric variants

The data therefore suggest that TTNtv are associated with lower LVMi, after adjusting for

other predictors of LVMi in this cohort. However, the original analysis included patients with

rare variants in other genes encoding sarcomeric proteins, which could have a similar or

disparate effect on LV mass. The analysis was therefore repeated, including building of the

baseline model predicting LVMi and then adding TTNtv to the optimized model, after

excluding 41 patients with rare non-truncating variants in MYH7 (n=29) and TNNT2 (n=11)

(1 patient had a rare variant in both genes) (note, no formal ACMG interpretation of

pathogenicity was performed).

In this subset cohort (n=675), gender and hypertension remained the predictors of LVMi. On

addition of TTNtv to this model, TTNtv remained associated with lower LVMi with little

change to the original estimate or p value (Table 3-12).

Table 3-12: Sarcomeric variant sensitivity analysis. Results of multivariable regression analysis

evaluating predictors of left ventricular mass in a cohort excluding individuals with rare variants in

MYH7 or TNNT2.

Variable Estimate change in LVMi (g/m2) P value Lower CI Upper CI

Male gender 17.3 <0.0001 13.4 21.2

History of hypertension 5.8 0.005 1.8 9.9

Presence of TTNtv -7.5 0.01 -13.2 -1.8

203 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.5.4 TTNtv positional effect

In previous pilot work from our group (n=42), it appeared that TTNtv that affected the

protein distally towards the C’ terminus were associated with more severe cardiac

endophenotypes. This was therefore evaluated further in this larger cohort. In Figure 3-15,

each CMR phenotype is plotted against the normalized cDNA position of the TTNtv

(individual cDNA position divided by 107976, the total cDNA length of the meta transcript).

When evaluating LVEF, RVEF, LVMi and biventricular volumes, there was no significant

linear relationship between TTNtv location and severity of the cardiac phenotype suggesting

that in this cohort there is no evidence to support a positional dependent phenotypic severity

effect with TTNtv (Figure 3-15).

204 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-15: Plots show the relationship between TTNtv location and cardiac endophenotypes assessed by

CMR in the cohort (n=83). The normalized cDNA position of the TTNtv on the axis is represented as the

position relative to the amino (N) and carboxyl (C) termini of the titin protein. Only variants in

constitutive exons are plotted. A regression line with 95% confidence intervals is shown for each variable

and slope and p value of the regression shown above each plot. LV/RV=left/right ventricle; EDVi, indexed

end-diastolic volume (mL/m2); ESVi, indexed end-systolic volume (mL/m2); SVi, indexed stroke volume

(mL/m2); EF, ejection fraction (%).

Whilst there was no positional severity effect, it appeared that the majority of variants in the

DCM cohort (although crucially not exclusively) clustered towards the distal two thirds of the

protein. A recent zebrafish study identified an internal promoter in the distal I band (just

before exon 240) for a short titin isoform, named Cronos, active in mouse and human hearts,

205 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

which when disrupted resulted in a severe phenotype(331). The hypothesis proposed was that

the Cronos isoform could partially rescue N-terminal truncations and that disruption of both

the full-length TTN and Cronos protein products results in more severe disease than

disruption of the full-length titin protein alone. Replotting the position-phenotype graphs,

annotated with the position of the Cronos promoter, shows a marked preponderance of

variants distal to this location, but no phenotypic distinction (Figure 3-16).

Figure 3-16: Cronos: Plots show the relationship between TTNtv location and cardiac endophenotypes

assessed by CMR in the cohort described in Figure 3-15, annotated by the expected position of the Cronos

promoter. Only variants in constitutive exons are plotted. The dashed line indicates the expected position

of the Cronos promoter (upstream of exon 240, numbered according to the LRG391_t1 meta-transcript).

206 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

In summary, these data show that there is marked heterogeneity in the cardiac phenotype of

TTNtv and that this is not fully accounted for by either variation in position, alternative

splicing (as all variants studied are in constitutively expressed exons) or by potential Cronos

disruption. This suggests that additional factors may modify expressivity, which I now

explore further.

3.4.6 Alcohol as a phenotypic modifier of TTNtv cardiomyopathy

3.4.6.1 Alcohol consumption

As part of the entry criteria to the study, no patient in the study met accepted criteria for a

diagnosis of alcoholic cardiomyopathy (>80g/alcohol per day for at least 5 years(332)). There

were 111 (16%) patients with a history of moderate excess alcohol consumption as defined in

the study design. Across the cohort, the mean weekly current alcohol consumption was 8.1 ±

15.9 units (IQR 0-10 units; range 0-200 units). The majority of the cohort reported that no

alcohol was consumed on a weekly basis (n=366, 51%) (Figure 3-17). There were 2

individuals with self reported consumption of alcohol of 189 and 200 units per week.

Figure 3-17: Self reported weekly alcohol consumption (units) across study cohort, stratified by gender.

207 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Patients with a history of alcohol excess were more likely to be male and have a lower RVEF

compared to patients without a history of alcohol excess (Table 3-13). On unadjusted

analysis, there was evidence of increased biventricular dilation and reduced left ventricular

function in patients with alcohol excess, but this was not significant after correcting for

multiple testing. There was no difference between groups in age at recruitment, family

history of DCM, left atrial volume, or presence of mid wall fibrosis.

208 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-13: Baseline demographics and cardiac phenotype, stratified by alcohol status. Continuous data

are summarised as mean (standard deviation) or median [interquartile range] and compared using

Mann-Whitney tests. Categorical variables are summarised as count and percentages and between group

differences tested with Fisher’s exact test. *Indicates significance after adjusting for multiple testing

(p≤0.003).

Alcohol Excess FALSE Alcohol Excess TRUE P value

N=605 N=111

Age at recruitment (years) 53.4 (14.5) 53.9 (13.1) 0.77

Male gender 365 (60.3) 104 (93.7) <0.001*

Family history DCM 97 (16.0) 16 (14.4) 0.78

Family history SCD 88 (14.5) 19 (17.1) 0.47

LVEF (%) 41.0 [30.0, 49.0] 37.0 [27.5, 47.0] 0.02

LVEDVi (mL/m2) 116.8 [101.9, 142.8] 123.0 [110.4, 144.6] 0.02

LVESVi (mL/m2) 67.5 [52.4, 96.0] 79.6 [58.7, 102.7] 0.009

LVSVi (mL/m2) 47.5 (13.5) 45.7 (12.2) 0.19

LVMi (g/m2) 85.2 [72.5, 102.6] 89.7 [79.8, 111.1] 0.007

RVEF (%) 54.0 [44.5, 62.0] 48.0 [39.0, 56.0] <0.001*

RVEDVi (mL/m2) 83.8 [69.4, 100.6] 91.2 [75.7, 102.9] 0.007

RVESVi (mL/m2) 38.8 [27.3, 51.9] 45.1 [36.5, 58.3] <0.001*

RVSVi (mL/m2) 44.0 (12.8) 43.2 (12.6) 0.54

LAVi (mL/m2) 53.9 [43.0, 68.1] 57.6 [46.8, 72.7] 0.05

Mid-wall fibrosis LGE 203 (33.6) 47 (42.3) 0.08

TTNtv present 70 (11.6) 13 (11.7) 1.00

209 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.6.2 Alcohol consumption and cardiac phenotype in patients with a truncating

variant in titin

Overall, amongst patients with a history of alcohol excess, 13 patients (16%) had a TTNtv

compared to 98 patients (15%) without TTNtv (p=1.00) (Table 3-13). Patients with TTNtv

and a history of alcohol excess had a mean LVEF of 27.7% compared to 39.8% in patients

with TTNtv without a history of alcohol excess and 37.8% in patients without TTNtv with a

history of alcohol excess (Figure 3-18). This suggested an additive effect of TTNtv and

alcohol on LVEF. To explore this further, a regression model was developed predicting

LVEF in this cohort and the significance of an interaction between TTNtv and alcohol was

evaluated.

Figure 3-18: LVEF in patients with TTNtv and alcohol excess. Beeswarm and overlaid boxplot showing

the distribution of LVEF (left ventricular ejection fraction) in cohort, stratified by TTNtv status and

alcohol history. Black bars are median values, blue squares indicate mean values. P value reported from

Kruskal-Wallis test compares LVEF across all 4 groups.

210 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.6.3 Predictors of LVEF

On univariable linear regression analysis, potential predictors of LVEF were evaluated. To

build the baseline model, before evaluating the importance of TTNtv and alcohol, biological

and clinical variables that could plausibly affect LVEF were selected. These were age,

gender, family history of DCM, the presence of left bundle branch block, the presence of

mid-wall fibrosis LGE, a history of atrial fibrillation and heart failure medication use (beta

blocker, ACE inhibitor and aldosterone antagonist). Medication was included to control for

the potential confounding effect of prognostic heart failure medication. Diuretic use or

NYHA class were not evaluated as predictors of LVEF as whilst they are associated with

LVEF severity, they do not contribute to LVEF impairment. RVEF was not evaluated as it is

highly correlated with LVEF (r=0.56, p <0.0001). The presence of mitral regurgitation was

also not evaluated as a predictive variable of LVEF because primary valvular disease was an

exclusion criteria, meaning that any mitral regurgitation reflected LV dilation and impairment

and was not a primary contributor.

Firstly univariable linear regression of each of the selected predictors against LVEF was

performed. Variables that were significant at a p value threshold of <0.10 were retained. Of

these variables, age, gender, family history of DCM, mid wall fibrosis LGE, history of AF,

and medication use were significant (Table 3-14).

211 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-14: Results of univariable and multivariable linear regression evaluating predictors of LVEF.

The adjusted analysis is adjusted for gender, family history of DCM, a history of atrial fibrillation, beta

blocker and aldosterone antagonist use and mid wall fibrosis. The process of model building is outlined in

the text.

Unadjusted analysis Adjusted analysis Variable Estimate 95% CI P value Estimate 95% CI P value change in change in LVEF (%) LVEF (%) Age -0.9 -1.5 to -0.2 0.01 - - - (per 10 years) Male gender -3.4 -5.4 to -1.5 0.0007 -2.3 -4.1 to -0.5 0.013 Family 7.1 4.5 to 9.7 <0.0001 5.7 3.4 to 8.0 <0.0001 history DCM Mid wall -3.7 -5.6 to -1.7 0.0002 -1.4 -3.1 to 0.4 0.13 fibrosis LGE LBBB -1.8 -3.9 to 0.4 0.11 - - - Atrial -4.6 -6.8 to -2.4 <0.0001 -3.3 -5.4 to -1.3 0.001 fibrillation Beta blocker -6.1 -8.1 to -4.1 <0.0001 -3.3 -5.2 to -1.4 0.0008 ACE -3.7 -6.0 to -1.3 0.002 - - - inhibitor Aldosterone -7.9 -9.8 to -6.0 <0.0001 -6.7 -8.6 to -4.9 <0.0001 antagonist

The full model was built including these 8 variables and then optimized the model using

reverse stepwise selection, with a p value threshold for exclusion of <0.05. At this stage, age,

LGE and the use of ACE-inhibitors were excluded.

Each discarded variable was added in turn (age, ACEi, LBBB, LGE) to assess if any variable

either became significant in the context of other variables, or whether their inclusion affected

the coefficient or significance of the existing variables. Of these, the inclusion of LGE

212 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

affected the gender variable significance value, though not the estimate. Therefore LGE was

retained, even though it was not a significant variable in the multivariable model (estimate

change in LVEF -1.4%, p= 0.13). On further analysis, 23% of female patients had LGE

compared to 41% of male patients (p<0.0001). The remaining discarded variables (age,

ACEi, LBBB) remained non significant and did not have an effect on the retained variables.

However, at this stage, given that age, a potentially important biological variable, was non

significant, and some of the included variables such as AF could be a reflection of LVEF (as

opposed to predictors of LVEF), subsets of the baseline model (termed Model 1a) were built

and compared using nested ANOVA testing. At this stage, I again tested the significance of

adding the previously discarded variables (Table 3-15). Models 1b-d were testing the variable

addition of age and ACEi compared to the baseline model, models 2a-2c were testing the

variable addition of LBBB, age, and LGE compared to the baseline model, and models 3a-3d

were testing the variable exclusion of medication compared to the baseline model. A simple

model of age and gender with the variable addition of family history of DCM and the

presence of LGE, whilst excluding medication use comprised models 4a-b. Finally, a model

excluding AF and medication use compared to the baseline model formed models 5a-5d.

The results, shown in Table 3-15, demonstrate that the baseline model was the best fit. The

addition of age, across all tested models, did not improve model fit. The exclusion of

medication and AF not did provide a better model for LVEF compared to the baseline model.

Therefore the final baseline model was Model 1a. The results of this adjusted model are

shown in Table 3-14.

213 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Table 3-15: Results of nested ANOVA comparing baseline models to predict LVEF.

Model Model variables Adjusted ANOVA ANOVA p number R2 testing value 1a: baseline Gender, family history DCM, AF, 0.1743 model Beta blocker, Aldosterone Antagonist, LGE 1b Gender, family history DCM, AF, 0.1709 Model 1a, 0.84 Beta blocker, Aldosterone Antagonist, Model 1b LGE, Age 1c Gender, family history DCM, AF, 0.1735 Model 1a, 0.62 Beta blocker, Aldosterone Antagonist, Model 1c LGE, ACEi 1d Gender, family history DCM, AF, 0.1702 Model 1a, 0.80 Beta blocker, Aldosterone Antagonist, Model 1d LGE, Age, ACEi 2a Gender, family history DCM, AF, 0.1738 Model 1a, 0.63 Beta blocker, Aldosterone Antagonist, Model 2a LGE, LBBB, Age 2b Gender, family history DCM, AF, 0.1732 Model 1a, 0.71 Beta blocker, Aldosterone Antagonist, Model 2b LGE, LBBB, Age, ACEi 2c Gender, family history DCM, AF, 0.1742 Model 1a, 0.54 Beta blocker, Aldosterone Antagonist, Model 2c LGE, LBBB, ACEi 3a Gender, family history DCM, AF, 0.1148 Model 1a, <0.0001 Beta blocker, LGE Model 3a 3b Gender, family history DCM, AF, 0.1637 Model 1a, 0.001 Aldosterone Antagonist, LGE Model 3b 3c Gender, family history DCM, AF, 0.1137 Model 1a, 1.0 Beta blocker, LGE, ACEi Model 3c 3d Gender, family history DCM, AF, 0.1629 Model 1a, 1.0 Aldosterone Antagonist, LGE, ACEi Model 3d 4a Age, gender 0.02688 Model 1a, <0.00001 Model 4a 4b Age, gender, family history DCM, 0.06588 Model 1a, <0.00001 LGE Model 4b 5a Gender, family history DCM, Beta 0.1623 Model 1a, 0.0006 blocker, Aldosterone Antagonist, LGE Model 5a 5b Gender, family history DCM, 0.1482 Model 1a, <0.00001 Aldosterone Antagonist, LGE Model 5b 5c Gender, family history DCM, Beta 0.1052 Model 1a, <0.00001 blocker, LGE Model 5c 5d Gender, family history DCM, LGE 0.06536 Model 1a, <0.00001 Model 5d

214 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.6.4 Effect of TTNtv and alcohol interaction on LVEF: Univariable analysis

The TTN*Alcohol excess interaction term was estimated to be associated with an additional

10.3% reduction in baseline LVEF compared to an individual with either TTNtv alone or a

history of alcohol excess alone (95% confidence interval -18.1 to -2.6, p=0.009; Table 3-16).

In the absence of other predictors of baseline LVEF, the effects of TTNtv and alcohol in a

patient with DCM compared to a patient with neither TTNtv nor a history of alcohol excess

can be summarised as follows (Table 3-16):

1) A patient with TTNtv alone and no history of alcohol excess has no statistically

significant change in LVEF (0.2% increase in LVEF, 95% confidence interval -2.8 to

3.3, p=0.87).

2) A patient with a history of alcohol excess but no TTNtv has no statistically significant

change in LVEF (1.8% decrease in LVEF, 95% confidence interval -4.4 to 0.9,

p=0.19).

3) A patient with both a history of TTNtv and alcohol excess has a 11.9% reduction in

LVEF (95% confidence interval -18.6 to -5.1, p=0.0006).

3.4.6.5 Effect of TTNtv and alcohol interaction on LVEF: Multivariable analysis

When the TTNtv*Alcohol excess interaction was added to the optimised regression model of

predictors of baseline LVEF, the effects of TTNtv and alcohol together in patients with DCM

remained a significant predictor of reduced LVEF over and above the presence of either

TTNtv or alcohol excess alone (Table 3-16).

In this optimised multivariable model, the TTN*Alcohol excess interaction term was

215 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

estimated to be associated with an additional 9.1% reduction in baseline LVEF compared to

an individual with either just TTNtv or just a history of alcohol excess (95% confidence

interval -16.2 to -2.0, p=0.01; Table 3-16). Therefore these data show that the presence of

TTNtv and a history of alcohol excess is estimated to be significantly different from having

either a TTNtv alone or a history of alcohol excess alone.

Table 3-16: Results of univariable and multivariable linear regression analysis evaluating the effect of

TTNtv and alcohol excess on LVEF. *indicates adjustment for variables in baseline Model 1a. § i.e. the

effect of TTNtv and alcohol excess compared to either TTNtv alone or alcohol excess alone.

Unadjusted analysis Adjusted analysis*

Variable Estimate 95% P value Estimate 95% P value (change in confidence (change in confidence LVEF, %) intervals LVEF, %) intervals Baseline: No 0 - - 0 - - TTNtv or alcohol excess TTNtv, no 0.2 -2.8 to 3.3 0.87 -0.6 -3.4 to 2.2 0.68 alcohol excess Alcohol excess, -1.8 -4.4 to 0.9 0.19 -0.3 -2.8 to 2.2 0.80 no TTNtv TTNtv and -11.9 -18.6 to -5.1 0.0006 -10.0 -16.3 to -3.8 0.002 alcohol excess

TTNtv*Alcohol -10.3 -18.1 to -2.6 0.009 -9.1 -16.2 to -2.0 0.01 excess interaction§

Adjusting for all variables predictive of baseline LVEF in this cohort, compared to a patient

without TTNtv or alcohol excess, the presence of both TTNtv and alcohol excess is

216 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

associated with a 10.0% reduction in LVEF (95% confidence interval -16.3 to -3.8, p=0.002;

Table 3-16; Figure 3-19). The magnitude of effect is similar to the effect of the interaction

term due to the negligible individual effects of TTNtv and alcohol excess on LVEF.

Figure 3-19: Forest plot showing results of final linear regression model predicting LVEF and the effects

of TTNtv and alcohol. Alcohol XS=alcohol excess, LGE=late gadolinium enhancement mid wall fibrosis,

FHx =family history, AF=atrial fibrillation, A-Antag=aldosterone antagonist.

217 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.6.6 Effect of TTNtv and alcohol interaction on LVEF: Model checking

Figure 3-20: Model checking of final linear regression model predicting LVEF with the TTNtv and

alcohol interaction. The Residuals vs Fitted plot shows that the residuals are equally spread above and

below the horizontal red line, without distinct patterns. This suggests that there are no underlying non-

linear relationships that have not been accounted for. The QQ plot shows whether the residuals are

normally distributed, and here the residuals appear to fit the expected distribution. In the Scale-Location

plot (square rooted standardized residual vs. predicted value, measure of homoscedasticity), the residuals

appear to be spread equally across a range of predictors until approximately 45%. After this point, the

residuals do not appear to have equal variance. However as the residuals vs fitted plot appears adequate,

the model appears acceptable. In the final Residuals vs Leverage plot, checking for influential

observations, there are no points outside of Cook’s distance (dashed red line not seen on plot). Points 164,

349 and 25 are discussed in the text.

The adjusted R2 of the final model (with TTNtv and alcohol) was 0.1824. The LVEF model

evaluating TTNtv and alcohol is significantly different to the LVEF model without TTNtv

and alcohol (nested ANOVA, p=0.02).

218 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

The model demonstrates an important gene-environmental interaction between TTNtv and

alcohol. To assess the validity of the model, I evaluated whether assumptions of linearity and

homoscedasticity have been met and potential influential outliers identified (Figure 3-20).

These plots show that the model is acceptable. Data rows 164, 349 and 25 were investigated

to check for potential data entry errors and none were identified. They represented 3 males;

one individual with TTNtv with a history of alcohol excess with LVEF 56%, and two

individuals with TTNtv without a history of alcohol excess with LVEF 11% and 12%.

Therefore these individuals did not fit the model of TTNtv plus alcohol results in worse

LVEF. However, the overall model fit was acceptable and there were no errors in data

capture, therefore all these individuals were retained in the final model.

3.4.6.7 Evaluating alcohol as a quantitative trait

The model presented evaluated alcohol excess as a binary trait. Given the small number of

individuals with TTNtv and alcohol, there is a danger of model overfitting. To mitigate this,

evaluating alcohol as a continuous variable might be more statistically powerful, as alcohol

data would be available on all TTNtv individuals in the cohort, not simply collapsed into

n=13 positive cases.

However, the evaluation of alcohol consumption as a quantitative trait was not appropriate

for 2 reasons:

1. Current self reported alcohol consumption at the time of recruitment may not have

accurately reflected a previous history of alcohol excess, particularly relevant for

those individuals who may have reduced alcohol intake on medical advice following

the diagnosis of DCM. [NB: Therefore in the primary analysis, medical records were

also reviewed to identify a history of alcohol excess].

219 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

2. The distribution of quantitative alcohol consumption was skewed and did not meet the

assumption of linearity required of linear regression modeling. This could however

have been partly addressed by transforming alcohol consumption as demonstrated

below:

log2(alcohol consumption +1)

Figure 3-21: Log2 transformation of alcohol consumption plotted against LVEF. Red dots represent

patients with TTNtv.

However, the large number of individuals with an appropriate value of 0 (i.e. not representing

missing data) limits the utility of this model. I then explored categorizing alcohol

consumption into 4 groups (no alcohol, 1-10 units, 11-20 units, >20 units per week).

However, this showed that there was no linear relationship between either these categories

and LVEF (Figure 3-22) or the TTN:alcohol interaction term of these categories and LVEF

(Figure 3-23). Together, these data confirm that it is not appropriate to evaluate alcohol as a

quantitative trait in this cohort.

220 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

Figure 3-22: Weekly alcohol consumption grouped into categorical variables and plotted against LVEF.

Figure 3-23: Forest plot of estimate and 95% confidence intervals of TTNtv:Alcohol interaction term

predicting LVEF, with alcohol stratified by categorical variables outlined in Figure 3-22. Each level

compared to baseline of TTNtv: No alcohol. The results show that there is not a linear relationship

between increasing categories of alcohol consumption and LVEF.

221 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.4.7 Other modifiers of TTNtv cardiomyopathy

3.4.7.1 Evaluation of hypertension as a phenotypic modifier of titin cardiomyopathy

The interaction between excess alcohol and hypertension is long established, showing

elevated blood pressure with excess consumption and a dose dependent reduction in blood

pressure with alcohol reduction(333,334). Therefore, it is possible that the alcohol-titin

interaction is either mediated via the effects of hypertension or actually reflects a

hypertension-titin interaction.

To address this, I evaluated hypertension as a possible phenotypic modifier of titin

cardiomyopathy. In this cohort, 213 patients had a history of hypertension, of whom 10 had

TTNtv and 33 had a history of alcohol excess. Amongst patients with TTNtv, there was no

difference in LVEF between patients with and without a history of hypertension (TTNtv

patients, LVEF (%): mean ± sd, HT+ 37.0 ± 14.8; HT- 37.9 ±13.8, p=1.0) (Figure 3-24).

Figure 3-24: LVEF stratified by TTNtv and hypertension status. Groups are compared using the

Kruskal-Wallis test.

222 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

On multivariable linear regression analysis controlling for variables predictive of LVEF (as

built previously), the presence of both TTNtv and hypertension together was not associated

with a change in LVEF (estimate change in LVEF 0.9%, 95% CI -6.2 to 8.0, p=0.82) (Figure

3-25). Therefore, there is no evidence that a history of hypertension is a phenotypic modifier

of titin cardiomyopathy.

Figure 3-25: Forest plot showing results of final linear regression model predicting LVEF and the effects

of TTNtv and hypertension. LGE=late gadolinium enhancement mid wall fibrosis, FHx =family history,

AF=atrial fibrillation, A-Antag=aldosterone antagonist.

3.4.7.2 Additional genetic modifiers of the titin cardiomyopathy phenotype

A comprehensive review of additional genetic variants modulating the titin cardiomyopathy

phenotype in this cohort is beyond the scope of this chapter as this would require a study of

common genetic variants. In terms of additional rare genetic variants in DCM genes in this

cohort, amongst patients with TTNtv (n=83), one had an additional variant in MYH7, one had

an additional variant in BAG3 and 3 had an additional variant in RBM20. The phenotype and

age at study recruitment of these patients is shown in Figure 3-26, compared to TTNtv

223 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

patients without additional variants. The numbers of patients with additional variants

precludes further analysis of the effect of genetic modifiers on the titin cardiomyopathy

phenotype.

Figure 3-26: LVEF and age at study recruitment in patients with TTNtv, coloured by the presence or

absence of additional rare genetic variants. In this cohort, 5 TTNtv patients had additional variants in

BAG3, MYH7 or RBM20. LVEF=left ventricular ejection fraction.

224 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.5 Discussion

Truncating variants in titin are the commonest single genetic cause of DCM(5,79,84). DCM

affects up to 1 in 250 people worldwide and despite optimal therapy, 5 year mortality

approaches 20%(12,13). Therefore, improved understanding of the clinical manifestations

and phenotypic drivers of TTNtv cardiomyopathy offers the scope for genetic based risk

stratification of patients with DCM to improve clinical outcomes.

What this study shows: Overview

In this study of 716 patients with DCM, we used the power of informed variant curation and

detailed cardiac phenotyping with CMR to define novel genotype-phenotype correlations in

TTNtv cardiomyopathy. We demonstrate that DCM due to TTNtv is associated with a

blunted hypertrophic response, highlighting possible disease mechanisms. We show that

TTNtv DCM is not associated with a distinct functional cardiac phenotype compared to non

TTNtv DCM but identify a gene-environmental interaction between TTNtv and alcohol,

suggestive of alcohol as a phenotypic modifier of TTNtv cardiomyopathy.

Cardiac phenotype of TTNtv DCM: limited hypertrophic response

In this study we demonstrated that TTNtv DCM is associated with thinner LV walls, and

lower LV mass, in the absence of significant differences in LV dilatation, after controlling for

important clinical covariables. This suggests that TTNtv is associated with a limited

hypertrophic response. Whilst the magnitude of wall thickness and LV mass differences are

not clinically informative and cannot be used to predict genotype based on phenotype, they

may offer a potential insight into the pathogenesis of TTNtv DCM. Notably, these findings

from human studies correlate well with recently published preclinical models of TTNtv

225 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

cardiomyopathy as outlined below(88).

At the sarcomeric level, titin plays a key role in the mechanotransductive response of the

cardiomyocyte and regulation of cardiac hypertrophy. Pathological hypertrophy, the presence

of cardiac dilation and cardiac dysfunction, develops in response to biological stressors, such

as neurohormonal activation or myocyte injury. It is thought to develop from an adaptive

compensatory response to a maladaptive state(335). The mammalian target of rapamycin

(mTOR) is an evolutionary conserved kinase that can modulate cardiac hypertrophy(336,337)

and increased mTORC1 signaling is an adaptive response to cardiac stress in rat models. In

patients with all cause DCM, elevated mTORC activation has been associated with a trend to

increased progression of heart failure(338).

It has recently been shown that rats with TTNtv had elevated mTORC1 signaling at baseline

(consistent with other cardiomyopathies) but critically were not able to increase adaptive

signaling when stressed(88). This may offer a biological explanation of the blunted

hypertrophic response observed in human TTNtv DCM, though the mechanistic link between

TTNtv and mTORC activation remains to be established and it is not yet known if this

aberrant mTORC activation is specific to titin cardiomyopathy(339).

Cardiac phenotype of TTNtv DCM: no evidence of positional dependent phenotypic

severity

In this study, we find that the functional cardiac phenotype in DCM patients with TTNtv is

not more severe than in patients without TTNtv. Left and right ventricular function was

similar in both groups. In this larger cohort compared to our previous pilot analysis on the

first tranche of this cohort, we did not replicate findings from our institution of a positional

226 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

dependent effect of TTNtv on cardiac phenotype. We showed that a more severely impaired

cardiac phenotype was not associated with more distal variants. This is consistent with recent

data in rat models and induced pluripotent stem cells (iPSCs) suggesting the mechanism of

action of TTNtv is haploinsufficiency, as opposed to a dominant negative protein, although it

is clear that the mutated allele is translated but not detectable on agarose gels(88,340).

Our data does not support a role of the Cronos isoform in modulating the TTNtv

cardiomyopathy phenotype. Zou and colleagues studied homozygous titin truncations in

zebrafish embryos and identified less skeletal muscle dysfunction in fish with Z disc and I

band truncations. They identified a novel titin isoform, Cronos, which has an internal

promoter in the distal I band. The authors proposed that mutations proximal to the promoter

were rescued by upregulation of the distal Cronos isoform, which could not happen with

distal mutations. However, in this cohort, whilst we see a marked clustering of variants distal

to the position of the internal promoter, there are clearly patients with DCM and TTNtv with

variants proximal to the promoter. Therefore the Cronos isoform hypothesis does not fully

account for the TTNtv DCM phenotype. We also do not identify a pattern of phenotypic

severity related to the position of the promoter. The variants that cluster just before the

promoter are not splice variants, predicted to affect the promoter. It is worth noting that in the

zebrafish study, levels of Cronos were highest in adolescence and declined in adult fish, and

the Cronos protein has yet to be identified in human hearts.

Cardiac phenotype of TTNtv DCM: support for environmental modifiers

We observed that there was a similar ratio of males to females in TTNtv DCM compared to

non TTNtv DCM. The gender distribution is interesting because DCM is known to have a 3:1

male:female ratio, yet one might expect an autosomal dominant inheritance genetic form of

227 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

DCM to have a 1:1 male to female ratio. This suggests that additional factors may be

important for the development of phenotype in individuals with TTNtv and the presence of

TTNtv alone may not be sufficient.

It is possible that female gender confers a protective effect in the presence of TTNtv, or that

males with TTNtv are susceptible to a greater extent to an additional genetic modifier and/or

have greater exposure/susceptibility to an environmental modifier. This latter hypothesis was

explored further when evaluating environmental phenotypic modifiers, with a particular focus

on alcohol.

Potentially pathogenic TTNtv are found in ~0.5% the general population (15,88), yet the

prevalence of TTNtv DCM in the population is ~1 in 2000 (DCM prevalence 1:250, with

~11% having TTNtv), indicating that not all individuals with TTNtv develop DCM. Pedigree

studies demonstrate variable expressivity in families with TTNtv. Preclinical models

demonstrate the development of a TTNtv phenotype only upon stress(94,341). These findings

are strongly suggestive of the 2nd hit hypothesis, requiring additional genetic or

environmental modifiers to develop a cardiac phenotype in the presence of TTNtv. To the

best of our knowledge, this is the first study to demonstrate an important gene-environmental

interaction between titin and alcohol.

We demonstrate that a history of at least moderate excess alcohol consumption, above UK

government weekly ‘sensible limits’ but well below the threshold defined for alcoholic

cardiomyopathy(97), is a modifier of the TTNtv phenotype in DCM patients. In the presence

of a history of TTNtv and alcohol excess, a patient with DCM has a 10.0% reduction in

LVEF compared to a patient without TTNtv or alcohol excess. Whilst the number of patients

228 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

with TTNtv and alcohol excess were small, the effect size is large and statistically significant,

suggesting that these observations have not occurred by chance.

The biological basis of the titin-alcohol interaction is unclear. Alcoholic cardiomyopathy has

long been linked to chronic excess alcohol abuse(342). It is thought to develop as a result of

oxidative stress, apoptotic cell death, impaired mitochondrial stress, perturbations in fatty

acid metabolism and accelerated protein catabolism(101). Crucially however, the mechanism

of the primary injury which initiates these changes is unknown. Two leading theories

implicate the alcohol metabolite, acetaldehyde, as directly toxic to cardiomyocytes, and

propose that alcohol directly leads to apoptosis(342-344). These theories however cannot

explain the mechanism by which alcohol cessation enables reversal of cardiomyopathy in a

subset of patients.

A recent study offers insight into the titin-alcohol synergy seen in our data. The authors

hypothesized that chronic alcohol exposure (8 weeks of ethanol vapour) would exacerbate

cardiac remodelling due to volume overload, modeled in rats through aortocaval fistula

surgery(345). At the end of the exposure period, rats with volume overload and chronic

alcohol exhibited increased cardiac wall stress and reduced collagen I/III expression

compared to rats with volume overload alone. Rats with volume overload alone showed

increased collagen III synthesis in the acute remodelling phase. This points to an impaired

compensatory hypertrophic response in rats with volume overload and alcohol. The authors

proposed that alcohol abuse accelerated the decompensation of the stressed heart. In isolation

these findings are interesting, but together with our data that TTNtv are associated with a

blunted hypertrophic response, we can piece together a ‘double hit’ mechanism by which

TTNtv and alcohol excess together lead to an impaired compensatory hypertrophic response

229 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

to volume overload, resulting in a more impaired phenotype than would be expected from

either TTNtv or alcohol alone. Further studies in animal models with TTNtv and alcohol

excess are required to investigate this hypothesis further.

3.5.1 Strengths of this study

As noted in recent ACCF/AHA guidelines(346) and many expert review papers(311,347) a

major limitation of studies documenting risk in patients with DCM and heart failure is

suboptimal phenotyping. By leveraging the precision phenotyping with CMR we were able to

identify and robustly quantify the subtle phenotypic finding of a blunted hypertrophic

response and the interaction between TTNtv and alcohol. These findings may not have been

detected on other imaging modalities such as echocardiography due to the marked inter and

intra observer variability and reduced sensitivity compared to CMR. CMR is the recognized

gold standard for accurate quantification of cardiac chamber dimensions and function.

In addition to the detailed phenotyping, a major strength of this work is careful informed

variant curation. Our group has developed expertise in the understanding of TTNtv and were

the first to report the association of TTNtv with DCM in a large scale study(79) and the first

to report the importance of exon usage (PSI) in TTNtv phenotype(84). All TTNtv in this

cohort were rare variants in constitutively expressed exons and were confirmed through

Sanger sequencing or review of mapped reads.

Finally, a key strength of this work is the sample size. To the best of our knowledge, this is

the largest prospective cohort of patients with DCM confirmed by CMR to undergo genetic

analysis. The findings reported as a result of this analysis are novel and have improved our

understanding of TTNtv DCM and of DCM.

230 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

3.5.2 Limitations of this study

There are several potential limitations to this work. The study is single centre, though as a

major regional centre for CMR referrals, patients were recruited from over 30 regional

district general hospitals, alongside consecutive referrals from a dedicated Inherited

Cardiomyopathy Service at our tertiary centre. It is routine practice in our institution for all

patients with DCM to undergo CMR (including prior to device implantation). This broad

referral base should mitigate bias and it is to be noted our study population reflects a range of

phenotypic severities. However, the CMR requirement may bias against those with the most

severe phenotypes, for example those who presented with severe life threatening arrhythmias

for whom device implantation was indicated prior to CMR.

The passive ascertainment bias of alcohol consumption data needs to be acknowledged. This

data was accrued via patient interview and clinical records. A history of alcohol excess was

not reflected in current alcohol consumption levels, and data on time between alcohol

cessation and recruitment to the study and the duration of alcohol excess was not consistently

available. Proxy measures of alcohol consumption such as serum gamma glutamyltransferase

(GGT) were also unavailable across the cohort. However, these limitations reflect a real

world setting where such variation is inevitable.

The cohort of individuals with TTNtv comprised probands and a smaller number of affected

relatives. The genotype-phenotype correlation has the potential to be biased by this, as a

milder phenotype was demonstrated in affected relatives compared to the proband.

Furthermore, a family history of DCM was associated with a milder LVEF on multivariable

regression analysis. This could have been skewed by individuals with a milder phenotype due

to earlier disease detection on family screening. However, in reality the number of affected

231 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

relatives with TTNtv was 4, therefore this is unlikely to represent a major bias. The

seemingly protective effect of family history on LVEF was driven by probands.

Variable penetrance has the potential to limit genotype-phenotype correlations in studies that

begin with genotype information and then screen for phenotype. In this study, DCM was

diagnosed prior to genetic testing, therefore it was not limited by penetrance. However, this

did limit our ability to evaluate the penetrance of TTNtv which would have been valuable

information in the field of titin cardiomyopathy.

This study builds upon a previously published cohort of patients with TTNtv DCM(84). The

overlap between the two studies is 275 patients, of whom 32 have TTNtv. This study

therefore reports findings on a much enriched, novel cohort of DCM TTNtv patients,

addressing novel hypotheses.

Finally, whilst we have presented novel results, these findings have not yet been replicated in

an additional cohort of patients with DCM and TTNtv.

3.6 Summary

These data show the potential of genotype-phenotype studies in DCM to inform patient

stratification and provide mechanistic insight into disease pathogenesis. We have

demonstrated that truncating variants in the gene for the sarcomeric protein titin are

associated with a limited hypertrophic response that may inform our understanding of TTNtv

DCM pathogenesis. We also demonstrate that the phenotypic expression of TTNtv

cardiomyopathy is modified by even moderate excess alcohol consumption. These data may

232 Template by Friedman & Morgan 2014 Word Chapter 3: The clinical manifestations and phenotypic drivers of titin cardiomyopathy

provide guidance for the clinical management of both individuals with TTNtv DCM and

those found to have TTNtv prior to the onset of overt cardiomyopathy.

3.7 Outline of further work

• Genetic study in the context of high alcohol intake is currently underway; in

collaboration with a Spanish group, we are currently undertaking sequencing of

patients with alcoholic cardiomyopathy to evaluate the burden of TTNtv. Our

hypothesis is that TTNtv contribute to alcoholic cardiomyopathy.

• In Chapter 5, the phenotype data forms the foundation to prospectively evaluate

clinical outcome stratified by genetic characteristics.

3.8 Acknowledgements

All work in this chapter is my own except where credited below:

• Sequencing and Sanger confirmation of variants was performed by– Rachel Buchan,

Sam Wilkinson, Alicja Wilk, Will Midwinter.

• The customized in house bioinformatics pipeline was developed and run by – Nicky

Whiffin, Risha Govind, Roddy Walsh, Shibu John, Liz Edwards.

• Nicky Whiffin extracted the list of all protein altering variants in 928 DCM samples

from the in house cardiodb database.

• Cardiac MRI measurements of chamber volumes and dimensions was performed by

the CMR reporting fellow at the time of clinical study.

• Inga Voges performed all left atrial measurements.

• The BRU research nurses recruited all patients to this study.

• Statistical mentorship and advice was provided by Simon Newsome.

233 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4 IMAGING PREDICTORS OF

CARDIAC REMODELLING

4.1 Aims and hypotheses

The overall aim is to evaluate imaging predictors of remodelling in DCM. Specific

hypotheses are as follows:

Primary Hypothesis

- Baseline left ventricular contractile reserve in patients with recent onset DCM is

predictive of LV remodelling at 1 year.

Secondary Hypotheses

- Baseline circumferential and longitudinal myocardial strain, replacement myocardial

fibrosis, interstitial myocardial fibrosis, and RV contractile reserve in patients with recent

onset DCM are each predictive of LV remodelling at 1 year.

- An adequate baseline left ventricular contractile reserve in patients with DCM reflects a

lower level of interstitial myocardial fibrosis when compared to patients with poor

baseline left ventricular contractile reserve.

234 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.2 Background

In this chapter I evaluate imaging predictors of cardiac remodelling, with a specific focus on

the predictive effect of left ventricular (LV) contractile reserve as assessed by low dose

dobutamine stress on left ventricular ejection fraction after 1 year of follow up in patients

with DCM.

As previously outlined, the prognosis of patients with DCM is variable, with a 5 year

mortality rate of ~20%(12,13), yet a potential recovery rate of over 20%(348-351). Resting

LV indices have not however been consistently shown to accurately predict myocardial

remodelling(172,179,194,195,350,351).

Myocardial contractile reserve as a prognostic marker

Myocardial contractile reserve is the difference in LV function at rest and under stress, either

pharmacological or exercise. Several studies have shown that contractile reserve, determined

through dobutamine stress, is a strong prognostic variable, independent of resting ventricular

indices, even in patients with severely impaired LV function(172-179). None of these studies

used CMR as the imaging modality. In one study, 62 patients with LVEF <30% (mean age 48

years) underwent dobutamine infusion up to 10 µg/kg/minute, with assessment of LVEF by

radionuclide ventriculography. Mean change in LVEF was 9%, and on multivariate analysis,

only the change in LVEF at stress was predictive of 1 year transplant free survival. In

another prognostic study, 71 patients with DCM underwent exercise radionuclide

angiography and echocardiography. The change in LVEF between rest and peak exercise was

an independent predictor for survival (p=0.0002)(177). The largest study of the prognostic

significance of contractile reserve was performed in 186 DCM patients who were assessed

235 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

for echocardiography wall motion score index changes following high dose dobutamine

infusion(352). The investigators showed that changes in the wall motion score index were a

stronger prognostic indicator for cardiovascular death during follow up compared to baseline

LVEF(352).

A limited number of historic studies have also evaluated the potential prognostic significance

of right ventricular (RV) contractile reserve(353). For example in a study of 67 patients with

advanced heart failure who underwent exercise stress and radionuclide ventriculography, an

increase in RVEF over 35% at exercise was an independent predictor of event free survival

(p=0.01)(354). A subsequent study, however evaluating both RV and LV contractile reserve

together, has shown that LV contractile reserve may be a stronger prognostic predictor than

RV contractile reserve(355).

Myocardial contractile reserve and reverse remodelling

Aside from prognosis, echocardiography based studies have evaluated the ability of LVEF

change during dobutamine infusion at baseline to predict subsequent improvement in LVEF

in patients on medical therapy. In one study of 18 patients with DCM (mean age 53 years,

mean LVEF 28%), changes in LVEF during low-dose dobutamine stress correlated with an

improvement in LV function during medium term follow up (r = 0.74, p < 0.001), though

analysis was limited by the relatively small number of cases and lack of adjustment for

confounders(173). In another study of 22 patients with new onset DCM, the LVEF change in

response to high dose dobutamine infusion (30 µg/kg/min) was also associated with 6-month

follow up LVEF (r=0.84, p=0.0001)(174). Similarly, a SPECT (Single-photon emission

computed tomography) imaging study of 26 patients with DCM also showed that the change

in LVEF following low dose dobutamine infusion (10 µg/kg/min) was associated with

236 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

improved LVEF, with a cut off stress induced LVEF change of 7% associated with 84.7%

sensitivity and specificity for improvement in LVEF(175). Another study of 51 patients with

heart failure showed that myocardial contractile reserve assessed by exercise

echocardiography is also a predictor of LV reverse remodelling (LVRR) after cardiac

resynchronization therapy (CRT) (6.5% exercise-induced increase in LVEF was associated

with a 90% sensitivity and 86% specificity to predict response after 6 months of CRT) (356).

Whilst these studies all indicate that contractile reserve is predictive of remodelling, these

studies were performed in relatively small cohorts, with variable definitions of LVRR,

variable medical therapy and inclusion criteria, and limited statistical analyses.

Limitations of LVEF based myocardial contractile reserve assessment

The assessment of LV contractile reserve through assessment of LVEF change following

dobutamine stress is potentially limited as it is a load dependent measurement(357). To

mitigate this potential limitation, in this current study, we sought to assess myocardial strain

as an additional measurement of myocardial contractile reserve. Strain is a unitless measure

of tissue deformation. As the left ventricle contracts, the myocardium shortens in a

longitudinal and circumferential dimension (negative strain) and thickens or lengthens in the

radial direction (positive strain). Displacement encoding with stimulated echoes

(DENSE)(358,359) is an accurate and reproducible technique to measure myocardial strain,

but its clinical application is limited by long breath hold times or navigator gating which can

be impractical and time consuming. We therefore developed an accelerated cine DENSE

sequence to enable the evaluation of strain in a breath hold time suitable for patients with

DCM.

237 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Biological basis of contractile reserve

The biological basis of the dobutamine induced contractile reserve predicting myocardial

recovery is thought to relate to the degree of beta-adrenergic receptor downregulation and

desensitization(181). Increasing beta-receptor downregulation reflects progressive

deterioration in left ventricular function(181). Therefore, improved contractility during

dobutamine administration reflects preserved beta receptor function, which reflects the

potential to show improvement in LVEF. Crucially, the original study linking contractile

reserve to beta-receptors showed that beta receptor down regulation can occur in patients

with only mild-moderate ventricular dysfunction, demonstrating how contractile reserve may

have additional prognostic value over and above LVEF alone(181). This data is in line with

clinical studies showing a poor correlation between resting LV function and contractile

reserve, suggesting that the presence or absence of contractile reserve reflects an independent

pathophysiology which could determine the likelihood of myocardial remodelling(172).

Further support for the link between cardiac sympathetic nervous system function and

myocardial contractile reserve was provided by a study of 24 DCM patients who underwent

iodine-123- metaiodobenzylguanidine (123-I-MIBG) and invasive assessment of contractile

reserve through LV catheterization(360). 123-I-MIBG is an imaging tracer that shares similar

myocardial uptake, storage, and release mechanisms as norepinephrine in sympathetic nerve

terminals, so is used to evaluate sympathetic nervous function. Myocardial contractile reserve

measured as the percentage change in myocardial contractile function (rate of pressure LV

pressure rise in systole; dp/dtmax(361)) between rest and peak atrial pacing, was reduced in

patients with reduced sympathetic function(360). The authors proposed that abnormal

myocardial contractile reserve was secondary to depleted intramyocardial norepinephrine.

They also demonstrated that reduced 123I-MIBG accumulation was associated with down

238 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

regulation of SERCA2 mRNA expression, indicative of altered calcium handling. These

findings were supported by a further study of 46 DCM patients which showed reduced

myocardial contractile reserve in response to low dose dobutamine was related to reduced

myocardial expression of regulatory protein mRNAs, including beta-adrengeric receptor,

SERCA, and phospholamban, involved in (362).

However, sympathetic dysfunction and the degree of beta-receptor down-regulation may not

fully account for the improved LV function reported in patients with preserved contractile

reserve. Abnormal myocardial perfusion has been linked to reduced contractile reserve in

small studies(363,364). The extent of myocardial contractile reserve has also been inversely

linked to myocardial interstitial fibrosis measured on endomyocardial biopsy(365). This has

not yet been demonstrated in vivo, but techniques such as CMR T1 mapping and extracellular

volume fraction quantification may provide a valuable insight(153). It is plausible that

contractile reserve reflects an absence of myocardial interstitial fibrosis, thereby providing an

LV substrate with the potential to remodel. Therefore the direct in vivo measurement of

myocardial fibrosis may be a novel prognostic marker of remodelling.

Summary

In summary, baseline myocardial contractile reserve has been associated with both prognosis

and myocardial remodelling in response to therapy patients with DCM. However, historic

studies have used LVEF echocardiographic assessment of contractile reserve with only

limited exploration of the biological basis of contractile reserve. CMR offers gold-standard

evaluation of cardiac structure and function, as well as in depth tissue characterisation. This

study therefore aims to bridge the gap between historic studies of contractile reserve and the

increasing utility of CMR in DCM, with in vivo tissue characterisation permitting assessment

239 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

of replacement and interstitial myocardial fibrosis and precision quantification of cardiac

function, to identify imaging predictors of LV remodelling.

4.3 Methods

4.3.1 Study overview

The study was a prospective observational study, evaluating whether the imaging parameters

of contractile reserve, myocardial strain and myocardial fibrosis, measured at baseline in

patients with recent onset DCM, predict follow up LVEF at 1 year (Figure 4-1). Contractile

reserve was the primary imaging parameter evaluated.

Figure 4-1: DCM Remodelling study overview.

4.3.2 Study cohort, inclusion and exclusion criteria

4.3.2.1 Dilated cardiomyopathy cohort

The primary study cohort consists of prospectively recruited patients with DCM diagnosed

within the preceding 1 year. Inclusion criteria were a diagnosis of DCM, over the age of 18

years old, the presence of sinus rhythm, and the absence of a contraindication to CMR

scanning or dobutamine administration.

240 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

The diagnosis of DCM was made based on CMR evidence of left ventricular dilation and

systolic impairment with reference to age, gender, and body surface area adjusted

nomograms(116) as previously outlined (Section 2.3.1.1, p115). Exclusion criteria for DCM

included a history of uncontrolled systemic hypertension, coronary artery disease (>50%

stenosis in one or more major epicardial arteries or previous percutaneous coronary

intervention or coronary artery bypass grafting), chronic excess alcohol consumption meeting

criteria for alcoholic cardiomyopathy (>80g/day for more than 5 years(97)), systemic disease

known to cause DCM, pericardial disease, congenital heart disease, infiltrative disorders (e.g.

sarcoidosis), or significant primary valvular disease(8,306,307).

As the primary analysis was the evaluation of LVEF at baseline and upon incremental

administration of dobutamine, sinus rhythm was a strict inclusion criteria, to ensure accurate

assessment of cardiac volumes using CMR.

A contraindication to CMR at 3T included the presence of a pacemaker, defibrillator, or

pacing wires, metal implants (including cochlear or spinal implants, hydrocephalus shunts),

vascular clips, or foreign bodies or metal in the eye (checklist shown in Appendix Figure

8-1).

A contraindication to dobutamine included any known hypersensitivity to dobutamine, recent

acute myocardial infarction, a history of unstable angina, recent life threatening arrhythmia,

severe dynamic or fixed left ventricular outflow tract obstruction, active endocarditis,

myocarditis or pericarditis, serious systemic illness, NYHA symptoms class IV, known

moderate left main stem or proximal left anterior descending coronary artery stenosis, or

severe systemic hypertension (systolic blood pressure >180mmHg and/or diastolic blood

241 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

pressure >120mmHg).

Patients were recruited via referral from a network of regional Cardiologists, via recruitment

at the time of referral for CMR to our institution, or via self-referral from the

Cardiomyopathy UK website.

4.3.2.2 Healthy volunteer cohort

The primary analysis was evaluation of contractile reserve as a predictor of change in LVEF

within the DCM patient cohort. A cohort of healthy volunteers (HVOL) was recruited to

permit comparison of the baseline contractile reserve response in DCM patients and normal

subjects. These individuals had no history of medical illness, were not taking regular

medication, and did not have evidence of cardiac structural or functional impairment on CMR

scanning. Initially, they were recruited via advertisement within the hospital Trust. In order to

recruit more older males (to age match the controls), a subsequent major ethics amendment

was submitted to permit recruitment beyond the hospital, in local community sports centres,

libraries, community centres and churches.

All participants gave written informed consent and the study was approved by the relevant

regional research ethics committees.

4.3.3 Baseline CMR protocol

All patients underwent CMR at 3T (Siemens Skyra scanner). The scan protocol is shown

Table 4-1.

242 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-1 CMR baseline study protocol

DCM Remodelling CMR scan protocol

1. HASTE (Half Acquisition Single Shot Turbo-spin Echo) x 3 (transaxial, saggital, coronal)

2. Vertical long-axis (VLA), horizontal long axis (HLA), Left ventricular outflow tract (LVOT),

right ventricular outflow tract (RVOT)

3. Short axis (SAX) Cines

4. DENSE: 1 SAX, 2 long-axis (LAX) (HLA and VLA)

5. T1 mapping pre dobutamine (5,3,3)- basal and mid ventricular SAX slice (repeat x2)

6. Low-dose Dobutamine stress as an inotropic agent

Dose: 50mg made up to 50mL normal saline [concentration: 1mg/mL]

i. 5minutes at 5µg/kg/min

ii. Acquire SAX cines

iii. Further 5 minutes at 10 µg/kg/min

iv. Acquire DENSE: 1 SAX, 2 LAX

v. Acquire SAX cines (after 10 minutes of dobutamine infusion)

7. Wait until HR normalises

8. T1 mapping pre contrast– basal and mid ventricular SAX slice (Repeat x2)

9. Late Enhancement (Gadolinium 0.1mmol/kg)

10. T1 post contrast -15mins after Gd – basal and mid ventricular SAX slice (Repeat x2)

4.3.3.1 Assessment of ventricular volumes and function

Cardiac volumes were assessed 3 times: at baseline; following the administration of

dobutamine at 5µg/kg/minute; and at 10µg/kg/minute. Each time, end-expiratory breath hold

steady state free precession (SSFP) cine images were acquired in the 3 long axis planes

(horizontal long axis, HLA; right ventricular outflow tract, RVOT; left ventricular outflow

tract, LVOT) and 8mm short axis slices (2mm gap) from the atrioventricular ring to the apex

243 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

(13). At 3T (Skyra, Siemens), imaging parameters were slice thickness 8mm, time per

acquired cine frame / echo time (TR/TE) 2.8/1.2ms, acquired resolution typically 1.65

x1.65mm, flip angle typically 44 degrees, field of view 340x240mm2, acquired temporal

resolution 54ms and 25 reconstructed frames.

Left ventricular (LV) volumes, function and mass were measured using a semiautomated

threshold-based technique (CMRtools, Cardiovascular Imaging Solutions, London, UK)

(Figure 4-2).

Figure 4-2: Example of threshold based measurement of cardiac volumes using CMR tools. Orange= left

ventricle blood pool excluding papillary muscles; yellow= left ventricular myocardium; purple= right

ventricle blood pool.

Left and right atrial area (LAA, RAA) was measured by tracing the atrial endocardial borders

with exclusion of the pulmonary veins and LA appendage in the horizontal long axis view.

The mean of two LAA and RAA measurements was recorded. Septal and lateral left

ventricular wall thickness was measured at the level of the papillary muscles, mid cavity in

end-diastole. An average of at least 2 measurements was taken. Maximum left ventricular

wall thickness was measured in each short axis slice at end-diastole, with the exclusion of

244 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

papillary muscles. All volume and mass measurements were indexed to body surface area and

referenced to age and gender based tables(116).

4.3.3.2 Dobutamine assessment of contractile reserve

All patients were asked to stop beta blocker therapy for 48 hours prior to the scan. An ECG

was performed at baseline to exclude high degree atrio-ventricular block (second or third

degree heart block).

Figure 4-3: Schematic of dobutamine infusion stages. Baseline (Stage 0) is prior to the commencement of

dobutamine. Heart rate (HR) and blood pressure (BP) are recorded at baseline and every two minutes. At

Stage 1, infusion of low dose 5µg/kg/minute dobutamine is commenced and continued for at least 5

minutes, during which short and long-axis cines are acquired to assess cardiac volumes (Stage 1a). The

dobutamine infusion is then increased to 10µg/kg/minute (Stage 2), during which myocardial strain data

is acquired. After infusion of dobutamine 10µg/kg/minute for at least 5 minutes, short and long-axis cines

are acquired to assess cardiac volumes (Stage 2a).

245 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Following the first myocardial T1 map (see below), baseline heart rate and blood pressure

were recorded, then dobutamine infusion was commenced. An outline of the dobutamine

infusion protocol is shown in Figure 4-3. The dobutamine was administered via a peripheral

intravenous cannula with long line extension to the infusion pump located in the control

room. The concentration was 1mg/mL (50mg dobutamine in 50mL normal saline). Infusion

was commenced at 5 µg/kg/minute for 5 minutes (Stage 1). This very low dose ramped

protocol was administered to increase patient tolerability of the infusion. During this period,

cardiac volumes were assessed as outlined above (Stage 1a). The infusion was then increased

to the target dose of 10 µg/kg/minute for at least 5 minutes (Stage 2). At 2.5 minutes after

increasing the infusion, myocardial strain was assessed with the cine DENSE sequence. After

the 5 minute mark (at least 10 minutes of dobutamine infusion, Stage 2a), cardiac volumes

were again assessed. After the final SAX image, the dobutamine infusion was stopped.

Heart rate and non-invasive blood pressure were recorded at baseline, every 2 minutes during

dobutamine infusion, and after the infusion was stopped until heart rate normalized, blood

pressure normalized, and symptoms resolved. Patients were asked to report any symptoms at

regular intervals. In addition, all patients had an emergency alarm to press if they experienced

any untoward symptoms. Continuous ECG monitoring was performed throughout the scan,

including during the infusion of dobutamine. In addition to a cardiologist, a registered nurse

assigned to the study was present in the control room to assist in the event of any

complication. Intravenous beta blocker (metoprolol), and atropine was kept available in the

control room if needed, and access to full resuscitation equipment was always available. A

protocol for recognising and treating dobutamine complications was kept in the control room

for reference. Indications to terminate the dobutamine infusion were the development of

severe angina, dyspnea or other intolerable symptoms, a fall of systolic blood pressure more

246 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

than 20mmHg below baseline or more than 20mmHg below the previous stage, the

development of hypertension >240/120mmHg, serious arrhythmia including ventricular

fibrillation, ventricular tachycardia or supraventricular tachycardia, second or third degree

heart block, symptomatic bradycardia, ST segment elevation, physical signs of hypoperfusion

such as cyanosis or pallor, or cerebrovascular symptoms such as ataxia or presyncope.

4.3.3.3 Assessment of interstitial fibrosis

All patients underwent T1-mapping at basal and mid-ventricular short axis levels pre and 15

minutes post administration of a bolus dose of gadolinium-based contrast agent (Magnevist or

Gadovist, Bayer) (0.1mmol/kg). A shortened Modified Lock-Locker Imaging (MOLLI)

sequence(151) was acquired in 11 cardiac-cycle breath-holds. The acquired inversion-

recovery times and recovery periods within each breath-hold were designed to optimise

accuracy and precision for the typical T1 ranges expected pre and post gadolinium(153)

known as 5(3)3 and 4(1)3(1)2 respectively (these are the numbers of images acquired after

each inversion pulse, with recovery periods given in brackets). The T1-mapping package

provided by the scanner manufacturer automatically performs per-pixel (“pixelwise”) curve-

fitting with Look-Locker correction (366) to generate T1 “maps” (Siemens Medical Systems,

Erlangen, Germany; “Myomaps” product).

T1 maps were acquired both pre and post infusion of dobutamine (after heart rate returned to

baseline), to control for any potential effect that dobutamine could have on the accuracy of

measurement or on T1 itself.

The trigger time for the T1 mapping sequence was adjusted to image in the diastolic rest

period, determined based on short axis cine images.

247 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Imaging parameters for T1 mapping were: non-selective inversion pulses, steady state free

precession single shot imaging with 20 degree flip angle, field of view 360x308mm, at

acquired resolution (frequency encoding x phase encoding) 1.9x2.4mm, slice thickness 8mm,

repetition time/echo time (TR/TE) 2.4/1.0ms. To achieve a minimal single-shot image

duration of 159ms, two further acceleration methods were applied: parallel imaging with 24

fully sampled central phase-encoding lines, and partial phase-encode sampling (7/8ths) with

zero-filling before reconstruction.

Images were analysed using CMR tools (CMRtools, Cardiovascular Imaging Solutions,

London, UK). T1 values were measured in a narrow region of interest in the septum, taking

care to avoid blood pool contamination, and a circular region in the blood pool, taking care to

avoid papillary muscles. Each patient’s packed cell volume (haemtocrit: the cellular fraction

of blood) was measured on the same day as the scan and extracellular volume fraction (ECV

= interstitium and extracellular matrix) was calculated according to the formula(153):

1 1 − (!"#$ !"#$%&'$ !1 !"#) (!"#$%& !1 !"#) !"# = 1 − ℎ!"#!$%&'($ 1 1 !"#$ !"#$%&'$ !1 !"##$ − !"#$%& !1 !"##$

All myocardial T1mapping and ECV quantification was performed in line with the Society of

Cardiovascular Magnetic Resonance and European Society of Cardiology consensus

statement(367).

4.3.3.4 Assessment of myocardial strain using cine DENSE imaging

All patients underwent baseline assessment of myocardial strain using a modified cine spiral

DENSE sequence. Images were acquired at the mid-ventricular SAX level and 2 long axis

planes (horizontal and vertical) at rest and during administration of the 10 µg/kg/min

dobutamine dose.

248 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Image parameters for DENSE acquisition were: 3.2x3.2x8.0mm3 spatial resolution, 30ms

temporal resolution and 2 direction encoding at 0.06cycles/mm, slice thickness 8mm,

repetition time/echo time (TR/TE) 15/1.0ms. The field of view (FOV) was 224mm2 with a

breath hold duration of 14 RR intervals.

Images were analysed and strain extracted from the DENSE data using semi-automated

Matlab post-processing software from the University of Virginia (170,368,369). The first

stage of analysis was anatomical delineation, using either a contour or a region of interest

covering the LV in the imaged slice. For long axis images, this was done with a line contour

(Figure 4-4). For SAX images, both contour analysis and region of interest were defined, with

the region of interest manually defined between endo- and epicardial borders (Figure 4-5).

These contours were defined in either the peak systolic frame or a diastolic frame before

automated propagation to the other frames (motion guided segmentation), with subsequent

manual adjustment as required. Frames with significant artefact (at the start or end) were

discarded. Strain was then calculated in the segmented areas, generating regional polar-

strain/time curves for radial and circumferential strain, contour strain/time curves for

longitudinal strain in 2 planes and contour strain/time curves for short axis strain.

249 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-4: Examples of defining single line contour regions of interest for the assessment of long axis

strain. Panel A= horizontal long axis, panel B= vertical long axis. The contour lines are shown in orange.

The top images in each panel are magnitude images and the lower images are the phase images. The

displacement of each pixel is encoded in the image phase (x and y directions). The software uses this

phase data to track the displacement of the LV tissue throughout the cardiac cycle, relative to a reference

time, defined before the onset of systolic contraction. The software then uses the pixelwise displacements

to calculate myocardial strain.

250 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-5: Examples of defining regions of interest for the assessment of short axis strain. Panel A=

endo-epicardial contour for assessment of radial and circumferential strain, panel B= single line contour

for global short axis contour strain. The top images in each panel are magnitude images and the lower

images are the phase images. The displacement of each pixel is encoded in the image phase (x and y

directions). The software uses this phase data to track the displacement of the LV tissue throughout the

cardiac cycle, relative to a reference time, defined before the onset of systolic contraction. The software

then uses the pixelwise displacements to calculate myocardial strain.

4.3.3.4.1 Development of the modified cine DENSE sequence

A physicist made iterative modifications to a spiral cine DENSE sequence that had

previously been validated at 3T(168) in order to reduce the breath hold time from ~21

seconds to a suitable range for patients with DCM(370). The primary modification was to

decrease the field of view (FOV) through a novel technique called zonal excitation, which

uses selective radiofrequency pulses to image just the heart whilst avoiding wrap (Figure 4-6).

251 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Limit the field

of view in y

Limit the field

of view in x

Figure 4-6: Figure demonstrating zonal excitation. Modified pulse sequence showing use of in plane slice

select gradients to reduce field of view. The first radiofrequency (RF) pulse (purple arrow) limits the field

of view in the x direction to the small region between the purple lines shown on the SAX cine figure and

the second RF pulse (orange arrow) limits the field of view in the perpendicular direction, leaving a small

region of interest around the heart to be imaged. The schematic sequence diagram also shows the R wave

of the ECG used to trigger the acquisition, fat suppression RF pulse (fat sat.), the DENSE encoding and

spiral imaging gradients (blue) and the slice selective RF pulses (purple; RF(z)) with increasing flip angle

used for imaging. Figure courtesy of Andrew Scott (physicist)(370).

The biggest limitation encountered in this process was retaining enough signal to noise ratio

(SNR) for DENSE analysis as a FOV reduction occurs at the cost of a reduction in SNR.

Scanning at 3T partially helps to compensate for this loss of SNR. We also reduced the

bandwidth, akin to reducing the frequency of sampling across the spiral. Lower bandwidths

give higher SNR, though at the cost of longer spiral readouts which can result in image

blurring. This sequence also uses a variable flip angle to mitigate tag fade through the cardiac

cycle, increasing the diastolic SNR. A compromise was therefore achieved between SNR,

FOV size and breath hold duration. During the initial acquisitions used in optimising protocol

parameters we used navigator (non breath hold) acquisitions as reference data. At a ‘medium’

breath hold of 14 cardiac cycles, we achieve a spatial resolution of 3.2mm and a temporal

resolution of ~30ms.

252 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Initial data analysis of a pilot healthy volunteer cohort revealed problems with the phase

wrapping of the images, meaning that the data was not suitable for analysis. This was a

particular problem for the long axis images because the longitudinal motion of the heart at the

base, shown in the long axis planes, is greater than the radial or circumferential

displacements. To address this, following discussion with the developers of the DENSE

sequence we modified (Epstein Lab, University of Virginia), we switched from balanced

encoding to simple encoding. Balanced encoding is more time efficient at the expense of

more phase wrap as it encodes in diagonal directions. Simple encoding encodes

displacements in x and y directions only. High displacement encoding frequencies are

associated with phase wrapping. As a result, we also reduced the encoding frequency from

0.1cycles/mm to 0.06cycles/mm, again a compromise between being able to detect phase

changes corresponding to small displacements and avoiding phase wrap. We also found that

it was important to shim carefully to avoid off-resonance artefacts (which are particularly

problematic in spiral MRI techniques) and select appropriate receive coils in order to avoid

residual image wrap.

4.3.3.5 Assessment of replacement myocardial fibrosis

Renal function was checked in all patients prior to gadolinium contrast. In patients with

eGFR >30mL/min/1.73m2 late gadolinium enhanced (LGE) images were acquired using a

breath hold inversion recovery sequence following administration of 0.1mmol/kg of

gadolinium contrast agent (Gadovist, Bayer) and saline flush to ensure complete delivery,

with inversion times optimised to null normal myocardium. Images were acquired in 3 long

axis planes and all short axis levels corresponding to the cine images. Typical image

parameters were: slice thickness 8mm, flip angle 20°, repetition time/echo time (TR/TE)

6ms/1.54ms, acquired resolution (frequency x phase-encoding) 1.45x1.93mm, field of view

253 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

370mm x 308mm adjusted to suit body habitus or to minimise breathhold duration. All LGE

images were acquired a second time with a 90 degree rotation of the phase encode direction.

This assists in identifying true enhancement in the presence of artefact.

Magnitude reconstructions of the inversion recovery images were used, on the assumption

that only focal enhancement was relevant, which enabled the radiographer to adjust the

normal myocardium to near zero intensity, so that focal fibrosis was clearly identified by its

relatively bright signal (371). Mid-wall myocardial fibrosis was recorded as present if

detected in both phase-encoding directions and in 2 orthogonal views, with cross cuts taken

as appropriate.

4.3.4 Follow up imaging

After one year, patients underwent follow up imaging with CMR for the primary purpose of

evaluation of LVEF. For patients with a contraindication to CMR scanning, focused 3d

echocardiography for assessment of LV volumes was performed. Evaluation of follow up

LVEF was performed blinded to baseline scan results.

4.3.5 Sample size calculation and statistical analysis

Following statistical advice, the sample size calculation was performed using the

pwr.f2.test in R for linear regression models. For a power of 80% and a significance

level of 0.05, for univariable regression with an effect size of 0.33, a sample size of 24 is

required (pwr.f2.test(f2=0.33, sig.level = 0.05, power=0.8, u=1).

The effect size (f2) was calculated based on a R2 of 0.25 (f2=R2/(1-R2) for the relationship

between absolute contractile reserve (the change in LVEF between 0-10 µg/kg/min

254 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

dobutamine infusion) and follow up LVEF. This is a relatively conservative estimate, as the

previous echo studies showed a strong correlation between stress LVEF and follow up LVEF,

with r =0.7-0.8(173,372), which would give an f2 of >1, which would be considered a very

large effect size. Therefore, a more conservative estimate was used, to ensure that an

appropriate sample size was recruited and that the study did not remain underpowered should

such a large effect size not be observed. Allowing for a 10% drop out rate, target recruitment

was 26 patients with DCM.

The primary hypothesis would be evaluated through linear regression analysis, evaluating the

predictive capacity of absolute LV contractile reserve (difference between baseline LVEF

and maximum LV change after peak dobutamine stress) as a continuous independent variable

against the dependent variable of follow up LVEF. A specific cut off for contractile reserve

will not be applied (to indicate the presence/absence of contractile reserve). Relative LV

contractile reserve will also be evaluated (absolute contractile reserve/baseline LVEF). To

control for regression to the mean which can occur with repeated measurements on the same

individual, the regression model will be adjusted for the baseline LVEF (difference between

baseline LVEF and the sample mean of baseline LVEF), known as analysis of covariance

(ANCOVA)(373). Absolute contractile reserve will first be evaluated as a univariable

predictor and then in a multivariable analysis, adjusting for confounders and variables of

interest (Figure 4-7). Confounders affect both the exposure and outcome variable, whereas

variables of interest may cause extra variation in either variable (but do not affect both)

therefore adjusting for these variables may reduce the standard error of the exposure or

outcome variables. In pre-specified secondary analyses, RV contractile reserve, baseline long

and short axis myocardial strain, and contractile reserve assessed by peak global

circumferential and longitudinal myocardial strain will also be evaluated as predictors of LV

255 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

remodelling.

A control cohort was recruited to establish a descriptive comparator of baseline contractile

reserve between DCM patients and healthy individuals. Following statistical advice, target

recruitment was set at 20 patients. This was not the primary analysis. With 20 volunteers, we

would have 80% power, at an alpha level of 0.05, to detect a 9% difference in baseline

contractile reserve between patients and volunteers (using pwr.t.test in R).

A p value of <0.05 was considered significant and all analyses were conducted in the R

statistical environment (version 3.3.1).

Figure 4-7: Diagram outlining the relationship between the exposure variable (contractile reserve) and

the outcome variable (follow up LVEF), together with potential confounders (white circles - variables that

affect both the exposure and outcome variables) and variables of interest (blue and green circles- that

may affect either the exposure or outcome variables). Graph produced using DAGitty(374).

256 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.4 Results

4.4.1 Cohort size and loss to follow up

In total, 38 patients with an initial clinical and imaging diagnosis of DCM were recruited. All

diagnoses had been made by a Consultant Cardiologist in our institution or a regional

referring hospital. Of these, 31 patients had CMR evidence of DCM at the time of study

recruitment, completed the baseline study protocol, and were included in the primary

analysis. One patient had CMR evidence of DCM but we did not start the dobutamine stress

protocol due to the presence of large pleural effusions and the potential risk of

decompensation with stress. Three patients had initial echocardiographic evidence of LV

impairment (in the referring centre), but normalised LVEF on CMR by the time of

recruitment, so were not included in the primary analysis. A further 3 patients had an initial

echocardiographic diagnosis of DCM (made in the referring centre) but evidence of an

alternative diagnosis on CMR study (1 patient with ischaemic heart disease, extensive

territory full thickness myocardial infarction; 2 patients with marked left ventricular

hypertrophy not consistent with DCM).

Of the 31 patients in the primary analysis, 2 were lost to follow up. One patient developed

bladder cancer and withdrew from the study, the other patient suffered a family bereavement

and withdrew from the study. Therefore, there were 29 patients with DCM who completed

the study protocol (Figure 4-8). In addition, 21 healthy volunteers were enrolled as control

participants.

257 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-8: Overview of cohort recruitment and completion to follow up

4.4.2 Baseline demographics and CMR findings

Baseline demographics in the cohort are shown in Table 4-2. There was no significant

difference in age at enrollment (mean age (years) DCM 52.8 vs controls 46.1, p=0.06),

gender (DCM 72% male vs controls 67% male, p=0.76), or body surface area between

patients with DCM and healthy volunteers.

All healthy volunteers were asymptomatic. The majority of patients with DCM were NYHA

class I/II (n=28, 97%). Patients with DCM had a higher systolic blood pressure compared to

258 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

volunteers (123 vs 110mmHg, p=0.004), but no difference in resting heart rate (66 vs 61,

p=0.13). All patients were in sinus rhythm and six DCM patients (20.7%) had evidence of

left bundle branch block (LBBB) on surface ECG.

Amongst the DCM patients, 24 patients (83%) were on beta blocker therapy, 25 patients

(86%) were taking an ACE inhibitor or an Angiotensin 2 receptor blocker, and 17 patients

were taking an aldosterone antagonist (59%). Two patients had contraindications to beta

blocker therapy (asthma) and one patient could not tolerate beta blocker therapy due to

bradycardia. Two patients were taking ivabradine, one due to a contraindication to beta

blocker therapy. Nineteen patients (66%) were taking diuretics at the time of study

enrollment. Only one patient was not on any medication due to personal choice. The overlap

in medication is shown in Figure 4-9.

Figure 4-9: Baseline prognostic medication in DCM cohort. BB=beta blocker, ACEARB= ACE inhibitor

or angiotensin 2 receptor blocker, AA= aldosterone antagonist. One patient was not on any medication.

Two patients were also taking ivabradine (not shown).

As expected, DCM patients had higher indexed LV end diastolic and end systolic volumes,

higher indexed LV mass, and lower LV stroke volume and LVEF compared to healthy

259 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

volunteers (Table 4-2). There was no evidence of significant difference in RV end diastolic

volume between groups, though DCM patients had higher RV end systolic volume with

lower RV stroke volume and RVEF compared to controls (Table 4-2). There was no

difference in mean left or right atrial area between DCM patients and controls. DCM patients

had a higher maximal wall thickness and higher mean septal wall thickness compared to

controls (Table 4-2).

Amongst DCM patients, 1 patient was not given gadolinium contrast due to renal

impairment. Of the remaining, 14 patients (50%) had evidence of mid-wall fibrosis LGE

enhancement. No control participants had evidence of mid-wall LGE. In total, 9 DCM

patients (31%) had mild mitral regurgitation and 2 patients (7%) had moderate mitral

regurgitation on CMR study. There was no evidence of other valvular disease. One control

participant had minimal mitral regurgitation, otherwise no control subjects had evidence of

valvular disease.

260 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-2: Baseline demographics and CMR findings in cohort stratified by diagnosis. Continuous data

are shown as mean (± standard deviation) and compared using the Mann-Whitney test, categorical data

are shown as count (percentages) and compared using Fisher’s exact test. LV/RV=left/right ventricular;

EF=ejection fraction; EDVi/ESVi=indexed end diastolic/end systolic volume; SVi=indexed stroke volume;

LVMi=indexed LV mass; LAA/RAA=left/right atrial area; ECV=extracellular volume fraction.

DCM HVOL P value N=29 N=21 Age at enrollment (years) 52.80 (12.1) 46.05 (11.9) 0.06 Gender = M (%) 21 (72.4) 14 (66.7) 0.76 Body surface area (m2) 2.00 (0.26) 1.91 (0.22) 0.20 LBBB 6 (20.7) 0 (0.0) 0.03 Systolic BP (mmHg) 123.2 (16.4) 110.1 (12.4) 0.004 Resting HR (bpm) 66.7 (15.2) 61.1 (8.1) 0.13 NYHA <0.001 1 14 (48.3) 21 (100.0) 2 14 (48.3) 0 (0.0) 3 1 (3.4) 0 (0.0) LVEDVi (mL/m2) 122.7 (24.9) 84.4 (10.4) <0.001 LVESVi (mL/m2) 74.3 (23.8) 28.5 (6.2) <0.001 LVSVi (mL/m2) 48.5 (12.6) 55.8 (7.9) 0.02 LVMi (g/m2) 79.9 (23.5) 58.8 (20.7) 0.002 LVEF (%) 40.2 (10.4) 66.3 (5.6) <0.001 RVEDVi (mL/m2) 94.6 (21.9) 90.5 (12.2) 0.44 RVESVi (mL/m2) 48.2 (15.3) 36.9 (8.6) 0.004 RVSVi (mL/m2) 46.4 (12.3) 53.7 (9.5) 0.03 RVEF (%) 49.4 (9.1) 59.4 (6.9) <0.001 Midwall LGE 14/28 (50) 0 (0) <0.001 Mean LAA (cm2) 25.7 (5.3) 23.7 (4.3) 0.16 Mean RAA (cm2) 23.8 (4.4) 24.6 (5.4) 0.56 Maximum LV wall thickness (mm) 11.4 (2.2) 9.4 (1.6) 0.001 Mean lateral wall thickness (mm) 6.5 (1.4) 5.9 (1.1) 0.11 Mean septal wall thickness (mm) 9.1 (1.6) 7.1 (1.1) <0.001 Basal septal ECV 0.29 (0.06) 0.25 (0.02) 0.003 Mid septal ECV 0.29 (0.05) 0.25 (0.02) 0.014

261 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.4.3 Safety of dobutamine

One patient did not complete the infusion protocol to 10 µg/kg/min due to hypertension

(systolic blood pressure >180mmHg), and only had a maximal dobutamine dose of 5

µg/kg/min. No other adverse incidents occurred during or after the administration of

dobutamine in the remaining patients.

4.4.4 Heart rate and blood pressure response to dobutamine stress

During low-dose dobutamine infusion, there was no evidence of a significant difference in

chronotropic response between DCM patients and healthy volunteers (Figure 4-10). DCM

patients appeared to have a reduced range of heart rate response compared to volunteers but

this difference was likely driven by one volunteer with a high heart rate response (Figure

4-10). During stress, DCM patients had a different systolic blood pressure response compared

to volunteers. All participants showed either a rise or fall in systolic blood pressure (SBP)

greater than 5mmHg. Five DCM patients had a fall in SBP during stress compared to only

one volunteer, though this was not related to the degree of LVEF impairment (Figure 4-11).

262 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-10: Maximum change in systolic blood pressure (SBP) and heart rate during dobutamine

infusion in DCM patients and healthy volunteers (HVOL). Black lines indicate median values. Groups are

compared using the Mann-Whitney test.

Figure 4-11: Maximal change in systolic blood pressure (SBP) and heart rate during dobutamine infusion

plotted against baseline LVEF in patients with DCM and healthy volunteers.

263 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.4.5 Baseline LV contractile reserve

Amongst DCM patients, the range of absolute LV myocardial contractile reserve (difference

in LVEF at peak dobutamine 10 µg/kg/min compared to baseline LVEF) was -9 units to +23

units change in LVEF (%), with mean change of 9.7 units and median of 11.0 units. Amongst

healthy volunteers, no individuals had a fall in LVEF during stress. The range of absolute

contractile reserve in volunteers was +1 to +20 units change in LVEF (%), with mean 10.1

units and median 10.0 units. There was no evidence of a statistically significant difference in

the absolute contractile reserve response between DCM patients and healthy volunteers

(p=0.98) (Figure 4-12).

Figure 4-12: Comparison of the absolute change in LVEF from baseline to peak dobutamine stress (10

µg/kg/min) in patients and volunteers (HVOL). Black lines indicate median values. Groups were

compared using the Mann-Whitney test. There was no evidence of a significant difference in absolute

contractile reserve between groups.

264 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Absolute contractile reserve was not correlated with the SBP response (r=0.10, p=0.49) or

baseline LVEF (r=0.03, p=0.81) (Figure 4-13). This SBP relation is important because it

suggests that the reduced contractile reserve was not secondary to insufficient stress, so there

were situations where both patients or controls demonstrated a physiological response to

stress (change in blood pressure) but without a change in LVEF.

Figure 4-13: Absolute contractile reserve. Plots showing no clear relationship between absolute

contractile reserve and the maximal change in (left hand panel) systolic blood pressure (SBP) during peak

dobutamine stress or (right hand panel) baseline LVEF.

Although there was no statistically significant difference in age between DCM patients and

controls, DCM patients were almost 7 years older than controls on average. Therefore further

analysis was performed to check for any possible effect of age on the contractile reserve

response. On univariable and multivariable linear regression analysis, there was no evidence

that age, gender, case/control status, baseline LVEF, or SBP response were predictors of the

absolute contractile reserve response (Table 4-3).

265 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-3: Linear regression models evaluating the effect of baseline variables on LV absolute contractile

reserve.

Unadjusted Adjusted Variable Change in Lower Upper P Change in Lower Upper P absolute CI CI value absolute CI CI value LV LV contractile contractile reserve reserve Age -0.3 -0.2 0.1 0.67 -0.01 -0.2 0.2 0.91 (per year) Male gender -3.2 -7.6 1.2 0.15 -3.8 -8.7 1.00 0.12 Control 0.3 -3.9 4.5 0.87 -0.4 -8.4 7.6 0.92 status Baseline 0.02 -0.1 0.1 0.81 -0.02 -0.3 0.2 0.9 LVEF (per 1%) Change in 0.04 -0.08 0.2 0.48 0.07 -0.08 0.2 0.35 SBP (per 1mmHg)

As shown in the regression analysis and illustrated in Figure 4-13, there was no significant

association baseline LVEF and absolute contractile reserve response in either patients or

volunteers. This is important for the ANCOVA analysis in which all regressions evaluating

the importance of contractile reserve are adjusted for baseline LVEF.

The distribution of relative contractile reserve (absolute contractile reserve/baseline LVEF) is

shown in Figure 4-14. The distribution is skewed by a DCM patient with a baseline LVEF of

11% who showed an LVEF increase to 24% on dobutamine stress.

266 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-14: Plot of relative LV contractile reserve in DCM patients and controls. Black bars indicate

median values. Groups are compared using the Mann-Whitney test.

4.4.6 Baseline LV strain

4.4.6.1 Feasibility of acquisition and analysis of strain data

Cine DENSE data was acquired in all participants at baseline and during peak stress. The data

was however only analyzable in a subset of participants. Baseline data from two volunteers

and peak stress data from two volunteers were not analysed. Baseline data from 3 DCM

patients and peak stress data from 5 DCM patients were not analysed. This was not

necessarily all strain analysis, but at least one of SAX contour, radial, circumferential, HLA

contour, VLA contour, at baseline or stress. Images were excluded from analysis because of

off resonance artefact or insufficient signal to noise ratio (SNR) to permit analysis. An

example of insufficient SNR is given in Figure 4-15. This was exacerbated in, but not limited

to, images acquired during peak stress and occurred in similar frequency in patients (1.4%)

and controls (0.9%). The loss of SNR was most likely a result of the compromise between

SNR, FOV size and breath hold time as outlined in the Methods.

267 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-15: Cine DENSE analysis. Example of a study in which cine DENSE data analysis was not

possible due to insufficient signal to noise ratio. In the upper panel (the magnitude images) of this

horizontal long axis analysis, the signal intensity is low and almost fades completely in the lateral wall

(orange arrow). In the lower panel (the phase images), the corresponding region in the phase data is very

noisy (white arrow), and the delineation between this and the blood pool (speckled area corresponding to

phase noise– black arrow) is poor.

4.4.6.2 Baseline strain results

Summary strain curves for the DCM and control cohorts are shown in Figure 4-16. The

following results were derived for each participant at baseline and stress: peak global HLA

contour strain, peak global VLA contour strain, peak global SAX contour strain, peak global

radial strain and peak global circumferential strain.

268 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-16: Summary strain-time curves for DCM and control cohort. The plots show the median

longitudinal (horizontal long axis, HLA; vertical long axis, VLA) strain, short axis (SAX) contour strain,

and radial and circumferential strain in healthy volunteers (blue curves) and the DCM cohort (red

curves). Curves are shown with the inter-quartile range (dashed lines). Peak strain is reduced in all

dimensions in DCM patients compared to controls.

269 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Compared to healthy volunteers, DCM patients had reduced peak global long and short axis

strain (Table 4-4).

Table 4-4: Peak global long axis (HLA, VLA) and SAX strain in DCM patients and volunteers . HVOL=

healthy volunteer. Data are shown as median [interquartile range]. Groups are compared using the

Mann-Whitney test.

DCM HVOL P value

VLA contour peak global strain -0.10 [-0.11, -0.06] -0.15 [-0.16, -0.14] <0.001

HLA contour peak global strain -0.09 [-0.11, -0.07] -0.16 [-0.16, -0.14] <0.001

SAX contour peak global strain -0.10 [-0.12, -0.07] -0.16 [-0.18, -0.14] <0.001

Radial peak global strain 0.18 [0.08, 0.28] 0.42 [0.37, 0.51] <0.001

Circumferential peak global strain -0.11 [-0.13, -0.08] -0.17 [-0.19, -0.17] <0.001

Across the cohort, there was a strong linear relationship between LVEF and contour strain

and LVEF and circumferential strain in DCM patients and healthy volunteers (Figure 4-17).

There was also a linear relationship between LVEF and radial strain in healthy volunteers,

though in DCM patients this could not be clearly established, particularly for patients with

severely impaired LVEF (Figure 4-17). This may reflect LV wall thinning in patients with

DCM, leading to inaccurate estimates of radial strain likely because there are fewer pixels

from which to derive the strains, therefore the effect of an erroneous blood pool pixel is

heightened. There was no linear relationship between mean lateral wall thickness and radial

strain (r=-0.30, p=0.15).

270 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-17: Relationship between left ventricular ejection fraction and contour strains (top 3 panels) and

circumferential and radial strains (bottom 2 panels). Univariable linear regression lines are plotted and

the slope and p value of the regression shown above the plot. Points are colour coded by diagnosis;

red=DCM, blue=healthy volunteer. LVEF=left ventricular ejection fraction.

271 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

The strength of correlation between measurements of strain and LVEF and mean septal and

lateral wall thickness is shown in Figure 4-18.

Figure 4-18: Correlation matrix showing the size of the correlation (from -1 to +1) between LVEF and

mean wall thickness (WT) and strain, in DCM patients only. Only correlations with a p value <0.05 are

displayed. The larger the circle and the more intense colour indicates a stronger correlation. LVEF is

strongly negatively correlated with circumferential, VLA, HLA and SAX contour strains and weak-

moderately positively correlated with radial strain. Lateral wall thickness is moderately correlated with

radial, circumferential and HLA contour strain estimates, whereas septal wall thickness is not correlated

with strain in DCM patients. Both horizontal and vertical long axis strain estimates are strongly

positively correlated. The SAX contour strain is strongly positively correlated with circumferential strain

in DCM patients. Together, this suggests that radial strain in this cohort is a less robust marker for the

assessment of LVEF function.

272 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.4.6.3 Contractile reserve as evaluated by assessment of strain

The absolute difference in baseline and stress peak global circumferential strain, SAX

contour strain, horizontal and long axis strains in DCM patients and healthy volunteers is

shown in Table 4-5. With the exception of absolute difference in horizontal long axis strain,

there was no difference in contractile reserve assessed by myocardial strain between DCM

patients and healthy volunteers. This suggests that long axis strain may be a subtle or earlier

marker of LV dysfunction in DCM patients compared to other estimates of myocardial strain.

Table 4-5: Absolute difference in peak global long axis (HLA, VLA) and SAX strain in DCM patients and

healthy volunteers (HVOL). Data are shown as median [inter-quartile range]. Groups are compared

using the Mann-Whitney test. Negative contour strains and negative circumferential strain indicate better

LV function, whereas positive radial strain indicates better LV function.

DCM HVOL P value

SAX contour -0.01 [-0.02, 0.01] -0.01 [-0.04, -0.00] 0.15

HLA contour 0.01 [-0.01, 0.01] -0.02 [-0.03, -0.01] 0.002

VLA contour -0.01 [-0.02, 0.01] -0.01 [-0.02, -0.00] 0.34

Radial 0.03 [-0.01, 0.13] 0.10 [-0.01, 0.21] 0.39

Circumferential -0.01 [-0.03, -0.00] -0.02 [-0.05, -0.01] 0.48

4.4.7 Baseline interstitial fibrosis

T1 measurements were taken pre and post dobutamine infusion, after heart rate had returned

to baseline levels. There was a strong correlation between native T1 measurements pre and

post dobutamine, both at basal (r=0.96, p<0.00001) and mid-ventricular (r=0.93, p<0.00001)

levels, suggesting that dobutamine did not affect T1 measurement at 3T.

273 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Extra-cellular volume fraction (ECV) was calculated from pre and post contrast T1

measurements at LV basal ventricular and mid-ventricular levels. The mean of two

measurements was taken. Whilst DCM patients had a higher overall ECV at LV basal and

mid-ventricular levels (Table 4-6), there was considerable overlap in ECV between DCM

patients and control participants (Figure 4-19).

Table 4-6: Extra-cellular volume fraction (ECV) in DCM patients and healthy volunteers (HVOL) at LV

basal and mid-ventricular levels. Data are expressed as mean (standard deviation) and compared using

the Mann-Whitney test.

DCM HVOL P value

N=29 N=21

LV Basal ECV 0.29 (0.06) 0.25 (0.02) 0.002

LV Mid ECV 0.29 (0.05) 0.25 (0.02) 0.006

Figure 4-19: Extracellular volume fraction (ECV) across cohort. Whilst healthy volunteers (HVOL) had a

lower ECV compared to DCM patients, the beeswarm and boxplots show overlap in LV basal (left) and

mid (right) ECV measurements. Black lines indicate median values The ECV values are a mean of two

measurements.

There was a strong correlation between basal and mid LV ECV estimates (r=0.93, p<0.0001),

suggesting that there is no regional distribution of interstitial fibrosis in the septum in DCM

274 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

and healthy individuals(Figure 4-20). There may be other regional differences in interstitial

fibrosis distribution that were not assessed in this study such as in the lateral wall or apex.

One DCM patient had a much higher ECV compared to the rest of the cohort (ECV >0.40).

This was a 56 year old male, NYHA class I, with severe LV impairment (LVEF 36%) and

septal mid-wall fibrosis. For all analyses, care was taken so that all regions of interest for T1

calculations were drawn away from any clear areas of replacement fibrosis or the blood pool.

Amongst patients and volunteers, there was a weak correlation between ECV and LVEF

(Figure 4-20).

Figure 4-20: Relationship between ECV and LVEF. Scatterplot showing [TOP PLOT] the strong

correlation between ECV measurements at the basal septal and mid septal level in DCM patients (blue

dots) and healthy volunteers (red dots) and the weak correlation between ECV and LVEF [LOWER

TWO PLOTS].

275 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.4.7.1 LVEF remodelling

After 1 year follow up, all patients underwent repeat evaluation of LV function. Seven

patients had contra-indications to CMR so had 3d echocardiography assessment of LV

function. Of these, 3 patients had CRT devices, 4 patients had ICD devices. The baseline and

follow up LVEF for each patient is shown in Figure 4-21. The median absolute change in

LVEF between baseline and follow up studies was 13.0% (range -1% to 47%). Seventeen

patients (58.6%) showed an absolute improvement of LVEF greater than 10 units. One

patient showed a marked improvement from baseline LVEF 11% to follow up LVEF 58%.

Three patients showed either no change or a deterioration in LVEF on follow up.

Figure 4-21: LVEF change at 1 year. Figure shows baseline LVEF (blue dots) and follow up LVEF (red

dots) after 1 year for each of the 29 DCM patients. The majority of patients showed improvement in

LVEF, with only 3 patients showing a deterioration in LVEF on follow up imaging.

276 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.4.8 Evaluating absolute contractile reserve as a predictor of LV

remodelling

Univariable analysis

Absolute LV contractile reserve was a significant predictor of the follow up LVEF, after

adjusting for baseline LVEF in ANCOVA regression analysis (each 1% increase in absolute

LVEF during dobutamine stress was associated with a 0.4% increase in follow up LVEF,

p=0.007, Table 4-7).

The correlation between absolute contractile reserve and follow up LVEF (without adjusting

for baseline LVEF) was r=0.52 (R2=0.28), p=0.004.

Next, potential confounder variables (age and gender) and variables of interest (baseline

prognostic medication, follow up imaging modality, CRT implantation) were evaluated.

Univariable ANCOVA regression was performed on these and additional variables of interest

that could potentially affect remodelling and the results are summarized in Table 4-7.

Of these variables, only gender was a significant predictor of follow up LVEF on univariable

analysis, with male gender being associated with a 7.5% reduction in follow up LVEF

(p=0.01, Table 4-7). NYHA class was associated with follow up LVEF but this was driven by

one patient in NYHA class 3 who showed an LVEF improvement of 22 units. Surprisingly,

age and prognostic medication (either as a binary variable or dose as a continuous variable)

did not predict LVEF remodelling.

On univariable analysis, the follow up imaging modality (3d echocardiography or CMR) did

277 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

not predict the follow up LVEF, however this will still be adjusted for in multivariable

analysis to control for any variation in follow up LVEF due to imaging modality and not

contractile reserve. Only 3 patients had CRT devices and this did not predict follow up

LVEF.

The presence of baseline mid-wall LGE enhancement, left bundle branch block (LBBB) and

mitral regurgitation did not predict follow up LVEF (Table 4-7).

278 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-7: Univariable ANCOVA regression evaluating predictors of follow up LVEF.

Variable Unit Estimate change to P value follow up LVEF (95% confidence interval) Absolute LV Per 1% change in LVEF 0.4 (0.12 to 0.73) 0.007 contractile reserve after 10 µg/kg/min dobutamine stress Age at enrollment Per 1 year 0.02 (-0.23 to 0.27) 0.89 Gender Male -7.5 (-13.3 to -1.8) 0.01 Number of prognostic Per each unit increase in the 0.5 (-2.5 to 3.6) 0.73 medications (BB, count of number of ACEi/A2RB, AA, medications Ivrabadine) Beta blocker use Present 2.1 (-5.4 to 9.6) 0.57 ACE inhibitor/A2RB Present -0.4 (-7.2 to 6.4) 0.91 use Aldosterone Antagonist Present 2.1 (-4.1 to 8.2) 0.49 use Beta blocker use Per 1mg increase in dose -0.06 (-1.7 to 1.6) 0.94 ACE inhibitor use Per 1mg increase in dose 0.17 (-0.54 to 0.88) 0.63 Aldosterone Antagonist Per 1mg increase in dose 0.08 (-0.47 to 0.62) 0.77 use Follow up imaging 3d Echocardiography -1.4 (-9.1 to 6.3) 0.71 modality CRT Implanted during follow up -0.8 (-10.8 to 9.2) 0.87 period LBBB Present 0.89 (-6.5 to 8.3) 0.81 NYHA class Class 2 compared to class I 1.9 (-3.6 to 7.5) 0.48 NYHA class Class 3 compared to class I 15.7 (1.0 to 30.4) 0.03 Hypertension History of hypertension -0.2 (-6.1 to 5.7) 0.94 Mid wall LGE Present -3.5 (-9.2 to 2.2) 0.22 Mitral regurgitation Mild -1.4 (-14.0 to 11.3) 0.82 Mitral regurgitation Moderate -2.4 (-8.7 to 3.9) 0.43

279 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Multivariable analysis

In multivariable analysis, adjusting for potential confounders and variables of interest,

absolute contractile reserve remained an independent predictor of follow up LVEF.

Multivariable models were built, first including only variables significant on univariable

analysis (Model 1: absolute contractile reserve, gender, NYHA class) and then also including

pre-specified variables of interest (Model 2: absolute contractile reserve, gender, NYHA

class, age, follow up imaging modality, prognostic medication use, and CRT use).

In the adjusted analyses, the effect size of absolute contractile reserve on follow up LVEF

was slightly reduced. Each 1 percentage unit increase in absolute contractile reserve was

associated with a 0.34 percentage unit increase in follow up LVEF in Model 1 (p=0.03), and

a 0.36 percentage unit increase in follow up LVEF in Model 2 (p=0.04) (Table 4-8). These

models are shown in Figure 4-22.

Of note, male gender, whilst having a large negative effect size on follow up LVEF, was not

significant in the multivariable models.

In summary therefore, in this cohort, absolute LV contractile reserve was the only identified

predictor of LVEF remodelling at 1 year follow up in ambulant patients with DCM.

280 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-8: Multivariable ANCOVA regression analysis evaluating absolute contractile reserve. Model 1

includes only variables significant on univariable analysis, Model 2 also includes pre-specified variables of

interest.

Model Estimate unit 95% confidence P value change in follow up interval LVEF (%) Model 1: Absolute CR, gender, NYHA class Absolute CR – per 1% absolute 0.34 0.03 to 0.65 0.03 increase in LVEF after dobutamine infusion Male gender -5.3 -11.0 to 0.46 0.07 NYHA class III 7.0 -7.2 to 21.2 0.32

Model 2: Absolute CR, gender, NYHA class, age, follow up imaging modality, CRT, prognostic medication use Absolute CR – per 1% increase in 0.36 0.01 to 0.72 0.04 LVEF after dobutamine infusion Male gender -6.3 -12.8 to 0.11 0.05 NYHA class III 5.6 -10.4 to 21.6 0.47 Age per 1 year increase -0.12 -0.38 to 0.15 0.36 3d echo follow up imaging modality -0.16 -9.0 to 8.7 0.97 CRT implantation 0.58 -10.8 to 12.0 0.92 Prognostic medication –Per each unit 1.43 -1.68 to 4.54 0.35 increase in the count of number of medications

281 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-22: Regression models evaluating absolute contractile reserve as a predictor of follow up LVEF,

adjusted for baseline LVEF. The left panel is the univariable model, the middle panel is Model 1 (absolute

contractile reserve, gender, NYHA class) and the right panel is Model 2 (absolute contractile reserve,

gender, age, NYHA class, prognostic medication, CRT use, follow up imaging modality). The slope of the

regression line and p value of the estimate is shown above each plot.

Sensitivity analysis

Contractile reserve estimates from the change in LVEF could be affected by factors which

affect LVEF, including heart rate variability and mitral regurgitation.

Adjusting for the maximal change in heart rate during dobutamine infusion, and the degree of

baseline mitral regurgitation, there was little change in the estimate of absolute contractile

reserve on follow up LVEF (Table 4-9). Heart rate variability and mitral regurgitation

themselves were not predictive of follow up LVEF. Nested ANOVA showed that there was

no significant difference between Model 1 (above) and Model 1 with the addition of these

variables (p=0.76).

282 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-9: Contractile reserve sensitivity analysis, showing the effect of controlling for heart rate

variability and mitral regurgitation on absolute contractile reserve as a predictor of follow up LVEF.

Variables Estimate unit 95% confidence P value change in follow interval up LVEF (%) Absolute CR – per 1% absolute 0.39 0.05 to 0.73 0.03 increase in LVEF after dobutamine infusion Δ HR during dobutamine 0.07 -0.22 to 0.36 0.63 Mitral regurgitation: Mild -0.9 -12.5 to 10.6 0.87 Mitral regurgitation: Severe -1.7 -7.5 to 4.1 0.55

4.4.8.1 Contractile reserve as assessed by longitudinal, circumferential and radial strain

The assessment of LV contractile reserve through assessment of LVEF change following

dobutamine stress is potentially limited as it is a load dependent measurement. To mitigate

this potential limitation, contractile reserve assessed by the difference in peak strain at

maximal dobutamine dose and baseline was also evaluated as a predictor of follow up LVEF.

As shown in Table 4-10, peak global circumferential and radial strain contractile reserve,

short axis contour strain contractile reserve and horizontal long axis longitudinal strain

contractile reserve were not associated with follow up LVEF. In contrast, longitudinal strain

contractile reserve, measured in the vertical long axis plane, was significantly associated with

follow up LVEF (-6.8 absolute % units change in follow up LVEF for each 0.1 unit increase

in vertical longitudinal axis strain reserve, p=0.03).

283 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Table 4-10: Contractile reserve assessed by myocardial strain as predictor of follow up LVEF.

Strain contractile Estimate change in 95% Confidence P value reserve follow up LVEF interval (absolute % units) for each 0.1 unit increase in strain Peak global 3.5 -7.1 to 14.2 0.49 circumferential Peak global radial -0.5 -2.8 to 1.7 0.63 Peak short axis -2.1 -5.9 to 1.8 0.28 contour Peak horizontal long -3.4 -8.4 to 1.6 0.17 axis Peak vertical long axis -6.8 -12.7 to -0.8 0.03

4.4.9 Other predictors of LV remodelling

Because of the skewed distribution (Figure 4-14), relative LV contractile reserve was not

evaluated as a predictor of LV remodelling. As pre-specified secondary analyses, RV

contractile reserve, myocardial strain, interstitial fibrosis and replacement fibrosis were also

evaluated as predictors of LV remodelling.

RV contractile reserve

Amongst DCM patients, the range of absolute RV contractile reserve (difference in RVEF at

peak dobutamine 10 µg/kg/min compared to baseline RVEF) was -17 units to +24 units

change in RVEF (%), with mean change of 4.2 units and median of 4.5 units. Amongst

healthy volunteers, the range of absolute contractile reserve in volunteers was -4 to +17 units

change in RVEF (%), with mean 8.0 units and median 10.0 units.

284 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Figure 4-23: Absolute RV contractile reserve in DCM patients compared to controls. Groups were

compared using the Mann-Whitney test. Black bars indicate median values. HVOL= healthy volunteers.

There was no evidence of a statistically significant difference in the absolute RV contractile

reserve response between DCM patients and healthy volunteers (p=0.47) (Figure 4-23). RV

contractile reserve was not correlated with baseline LVEF (r=0.13, p=0.35). RV contractile

reserve was highly correlated with LV contractile reserve (r=0.81, p<0.00001, Figure 4-24).

Figure 4-24: Right and left ventricular contractile reserve. Scatterplot with regression line and 95%

confidence intervals showing the strong correlation between absolute right and left ventricular contractile

reserve. The slope of the univariable regression line and p value are shown above the plot.

285 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

However, despite the strong correlation between RV and LV contractile reserve, on

univariable ANCOVA analysis absolute RV contractile reserve was not predictive of follow

up LVEF (estimate change in follow up LVEF for each 1 unit absolute increase in RVEF

after dobutamine infusion: 0.2 units, 95% CI -0.04 to 0.47, p=0.10).

In contrast to relative LV contractile reserve, the distribution of relative RV contractile

reserve (absolute RV contractile reserve/baseline RVEF) was not skewed (Figure 4-25). On

univariable ANCOVA analysis, relative RV contractile reserve was predictive of follow up

LVEF, although the borderline p value suggests that this predictor is not as robust as absolute

LV contractile reserve (estimate change in follow up LVEF for each 0.1 unit increase in

relative RV contractile reserve: 1.2 units, 95% CI 0.02 to 2.33, p=0.047).

Figure 4-25: Relative RV contractile reserve in DCM patients and healthy volunteers (HVOL). Relative

RV contractile reserve is the absolute RV contractile reserve after dobutamine infusion divided by the

baseline RVEF. Black bars indicate median values. Groups are compared using the Mann-Whitney test.

286 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Baseline long and short axis strain

On univariable ANCOVA analysis no baseline measure of strain, either in the longitudinal,

circumferential or radial dimensions, was predictive of follow up LVEF (Table 4-11).

Table 4-11: Results of ANCOVA regression analysis evaluating baseline indices of strain as predictors of

follow up LVEF.

Variable Estimate change in follow 95% Confidence interval P value up LVEF (absolute % units) for each 0.1 unit increase in strain VLA contour peak 6.2 -3.9 to 16.3 0.22 global strain

HLA contour peak -1.4 -6.4 to 3.6 0.58 global strain SAX contour peak -3.2 -11.5 to 5.1 0.43 global strain

Radial peak global 0.74 -1.1 to 2.6 0.42 strain Circumferential peak -5.4 -21.4 to 10.7 0.50 global strain

Replacement myocardial fibrosis

Amongst the 28 DCM patients who were given gadolinium contrast, 50% had evidence of

mid-wall myocardial enhancement. The majority was located in the septum (n=10, 71%). The

presence of mid-wall fibrosis was not a significant predictor of follow up LVEF in this cohort

(estimate 3.5% reduction in follow up LVEF, 95% CI -9.2 to 2.2%, p=0.22).

287 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Interstitial fibrosis

No ECV measurements were predictive of follow up LVEF (ECV basal: -4.3% change in

follow up LVEF per 0.1 unit change in ECV, 95% CI -9.3 to 0.7, p=0.09; ECV mid: -4.3%

change in follow up LVEF per 0.1 unit change in ECV, 95% CI -9.9 to 1.2, p=0.11).

4.4.10 Mechanistic basis of contractile reserve

Absolute contractile reserve was not correlated with interstitial fibrosis, measured either via

basal (r= -0.19, p=0.35) or mid LV ECV (r=-0.20, p=0.31). Whilst the individual with a very

high ECV also had a fall in absolute contractile reserve, for the remaining cohort, there was

no strong linear relationship between higher ECV and lower contractile reserve, as might be

expected if reduced contractile reserve was underpinned by interstitial fibrosis (Figure 4-26).

Most notably, individuals with a fall in contractile reserve have similar ECV values to

individuals with higher absolute contractile reserve values.

Figure 4-26: Contractile reserve and ECV. The graph shows that there is no clear relationship between

ECV (left y axis for basal and right y axis for mid LV levels) and absolute contractile reserve (% unit

change in LVEF after peak dobutamine infusion, x axis).

288 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.5 Discussion

This main finding of this prospective study is that LV contractile reserve, assessed by low-

dose dobutamine stress CMR, is the only independent predictor of LV remodelling in patients

with recent onset DCM, over and above age, gender, symptom status, prognostic medication

use and mid-wall replacement fibrosis.

The study also demonstrates that the degree of myocardial contractile reserve is not related to

the burden of myocardial interstitial fibrosis, providing insight into the mechanistic basis of

contractile reserve. In addition, the study demonstrates that contractile reserve assessed by

longitudinal strain and relative RV contractile reserve are predictive of LV remodelling, but

baseline myocardial strain, replacement myocardial fibrosis and interstitial fibrosis are not

predictive of LV remodelling.

These results prompt 3 notable points for discussion: 1) the capacity for remodelling in DCM,

2) the role of contractile reserve in predicting remodelling in DCM, and 3) the biological

basis of contractile reserve.

4.5.1.1.1 The capacity for remodelling in DCM

Whilst there is no universally accepted threshold to define left ventricular reverse

remodelling (LVRR), almost 60% of patients in this study showed an improvement of LVEF

greater than 10 percentage units. This is towards the upper limit of previous findings (25-

70%) (184-195) and confirms the capacity for apparent recovery in DCM. That the majority

of the cohort demonstrated improved LV function may reflect that this is a contemporary

cohort of DCM patients treated with optimal medical therapy.

289 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

This study also confirmed that the capacity for remodelling was not intuitive and cannot be

predicted based on resting LVEF alone. Improvement in LVEF was seen across the spectrum

of baseline LVEFs including those with initial severely reduced LVEFs. This was the basis to

evaluate contractile reserve as a predictor of LV remodelling in DCM.

4.5.1.1.2 The role of LV contractile reserve in predicting LV remodelling in DCM

In this study of 29 patients with DCM, a 1% absolute increase in LVEF during dobutamine

stress was associated with a 0.4% increase in 1 year follow up LVEF. Whilst contractile

reserve has previously been evaluated as a predictor of remodelling in echocardiography

studies, this is the first study to evaluate contractile reserve assessed by CMR in combination

with LGE CMR for the prediction of LV remodelling in DCM. This provided the opportunity

to evaluate contractile reserve in the context of novel DCM imaging biomarkers such as

replacement fibrosis.

Absolute LV contractile reserve was the only independent predictor of LV remodelling in this

cohort, over and above age, gender, symptom status, and mid-wall replacement fibrosis.

Importantly, in this study, the majority of patients were on appropriate medical therapy and

prognostic heart failure medication was not predictive of follow up LVEF.

This current study therefore demonstrates that predictors of outcome in DCM (e.g. mid-wall

replacement fibrosis) and predictors of remodelling in DCM are not necessarily

interchangeable, and at least for the 12 month follow up period of the study, contractile

reserve was the strongest imaging predictor of functional myocardial recovery.

There are plausible biological explanations as to why LGE was not a significant predictor of

290 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

remodelling in DCM. A previous study evaluating CMR predictors of remodelling in

ischaemic heart disease showed that, compared to LGE, low dose dobutamine stress was both

an incremental predictor of remodelling and could stand alone with the same predictive

capacity as LGE(375). This suggests that contractile reserve may reflect the global capacity

of the myocardium to remodel, for example preserved beta-adrenergic receptor sensitivity,

whereas LGE mid-wall fibrosis is a focal insult that does not affect remote myocardial

remodelling.

Of course it must be stressed that the current study was not formally powered to evaluate the

role of age, gender and mid-wall fibrosis as predictors of LV remodelling. Male gender in

particular was a significant predictor of adverse LV remodelling on univariable analysis but

did not remain significant on multivariable analysis. This may reflect sample size as the

magnitude of effect of male gender was similar in multivariable analysis compared to

univariable analysis. Further interrogation of this could form the basis of future studies.

These findings may be of significance to patients and clinicians to understand which patients

may remodel with appropriate medical therapy. At present, there are no integrated robust

predictors of recovery in DCM. One potential utility of these findings could be in guiding

decisions about the timing of device therapy, particularly in younger patients who may meet

criteria for ICD insertion. Whilst potentially life saving, there are procedural risks of ICD

implantation as well as long-term risks of multiple generator replacements in younger

patients. The presence of demonstrable contractile reserve could provide support for a

deferred implantation strategy to allow an adequate interval for possible recovery (current

guidelines recommend at least 3 months of medical therapy). Conversely, identification of a

patient with little potential for reverse remodelling could identify a patient in need of early

291 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

intensive therapy including prompt referral for advanced therapy. This of course requires

further evaluation in a prospective study. The data could also inform clinical trial design,

stratifying therapies by groups with different remodelling profiles.

Aside from the potential clinical utility of LV contractile reserve as a predictor of LV

remodelling, this finding could provide insight into mechanisms and potential treatment of

LV dysfunction. This however requires an understanding of the biological basis of contractile

reserve.

4.5.1.1.3 The biological basis of contractile reserve.

A recognized limitation of previous studies of LV remodelling is a failure to correlate

functional outcomes with molecular, cellular or histological findings(376). There is therefore

great interest in the exploring the biological basis of myocardial recovery. As contractile

reserve is a predictor of myocardial recovery, the biological basis of contractile reserve could

be informative of the mechanism of myocardial recovery.

There are several possible mechanistic explanations for contractile reserve. Primarily,

improved contractility during dobutamine stress is thought to reflect preserved beta-

adrenergic receptor function. Down-regulation of beta-receptors occurs with LV dysfunction,

irrespective of severity(181). In line with this, resting LVEF is not predictive of contractile

reserve(172). Reduced contractile reserve in vivo is linked to beta-receptor down-regulation,

even amongst patients with mildly impaired LV function(362). This is supported by our

study, in which the magnitude of the LV contractile reserve response was similar in patients

and controls and could not therefore be predicted based on LVEF alone.

Contractile reserve may therefore simply reflect beta-receptor upregulation or sensitisation,

292 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

leading to a myocardial substrate that is more likely to respond to prognostic therapy,

including beta-blockade(372,377). However, these studies also showed that contractile

reserve may also predict myocardial recovery in patients not receiving beta blocker therapy.

In addition, the basis for contractile reserve has also been linked to altered calcium handling

(reduced expression of SERCA2a and phosholamban mRNA)(362), suggesting that the

biological basis for contractile reserve is likely to reside in multiple molecular pathways.

Therefore, this study set out to evaluate one of the mechanisms of myocardial remodelling

that has also been implicated in the contractile reserve response, notably changes in the extra-

cellular matrix. Whilst molecular pathways defining reverse remodelling have not been

clearly defined, the transition from compensated to decompensated hypertrophy is better

understood and involves multiple molecular pathways, including re-expression of fetal genes,

impaired excitation-contraction coupling, vascular and cardiomyocyte growth mismatch,

myocyte necrosis, and changes in the extracellular matrix(378).

Interstitial fibrosis on endomyocardial biopsy has been inversely linked to the extent of LV

myocardial contractile reserve (365). Myocardial recovery is thought to be possible if there is

both a sufficient mass of viable myocytes and an absence of extensive fibrosis(379,380).

We therefore hypothesized that patients with a high ECV (an in vivo estimate of interstitial

fibrosis assessed through CMR T1 mapping) would have diminished contractile reserve.

However, this study showed that the magnitude of the contractile reserve response was not

related to the degree of interstitial fibrosis. Patients with both a rise and fall in LVEF during

dobutamine stress had a range of ECV values, showing that absolute contractile reserve is

unaffected by the burden of interstitial fibrosis. In addition, the magnitude of interstitial

293 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

fibrosis at baseline was not predictive of LV remodelling. This suggests that the interstitial

fibrosis previously noted to be correlated with contractile reserve may have simply been a

reflection of a failing myocardium and unrelated to contractile reserve or reverse cardiac

remodelling. Whilst interstitial fibrosis reflects dysfunctional myocardium, in this study, the

degree of interstitial fibrosis does not reflect the capacity of the myocardium to remodel.

4.5.1.1.4 Additional novel findings: the importance of longitudinal myocardial strain

In this study myocardial strain was assessed through the use of a novel modified cine DENSE

sequence. Previous cine DENSE sequences have been limited by long breath hold times but

in this study all patients were able to complete the breath hold required, demonstrating the

feasibility of evaluation of myocardial strain using cine DENSE CMR.

This enabled the assessment of circumferential, radial and longitudinal strain in DCM

patients. At baseline, DCM patients had reduced peak strain compared to controls, as

expected. However, the study also highlighted the importance of longitudinal function in

DCM patients.

At baseline, contractile reserve assessed by LVEF did not differ between DCM patients and

controls. Similarly, contractile reserve assessed by circumferential, radial and SAX contour

strain also did not differ between the DCM cohort and controls. However, contractile reserve

assessed by longitudinal strain (HLA) was reduced in DCM patients compared to controls,

suggesting that it is a more sensitive measure of LV dysfunction than LVEF. In line with

this, longitudinal strain contractile reserve (VLA) was predictive of follow up LVEF, which

has not previously been demonstrated by CMR assessment.

294 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

This replicates the finding that contractile reserve predicts LV remodelling in DCM and also

suggests that longitudinal strain is a good discriminator of LV function in DCM patients,

particularly amongst patients with a similar LVEF. Longitudinal strain measures different

aspects of myocardial deformation compared to LVEF, which largely reflects radial

contraction. Longitudinal strain is a measure of the active shortening of the LV in the

longitudinal direction and represents the function of subendocardial longitudinal myocardial

fibres. These fibres are more sensitive to reduced coronary perfusion and increased wall

stress, therefore longitudinal dysfunction may be an early marker of LV dysfunction(381).

The discrepancy between HLA and VLA measurements cannot be explained biologically, but

the direction of effect is that longitudinal strain assessment may be of value in patients with

DCM.

It is plausible that because global circumferential and radial strain were only measured at the

mid-papillary level, overall LV dysfunction may not have been adequately captured and this

is why these variables were not predictive of LV remodelling.

This finding of the importance of longitudinal strain is in line with a CMR dobutamine stress

study evaluating subclinical cardiac dysfunction in patients with liver cirrhosis and normal

LVEF compared to controls. The study found that only longitudinal strain differences on

intermediate dose dobutamine stress (and not circumferential or radial strain) discriminated

patients and controls(382). Echocardiography assessed longitudinal strain has been shown to

be of prognostic importance in patients with dilated cardiomyopathy(383), ischaemic heart

disease(384) and all-cause heart failure(385), attributed to it being a more sensitive measure

of LV dysfunction than LVEF. A recent retrospective echocardiography study of 96 patients

with LVEF <50% has also shown that baseline global longitudinal strain was moderately

295 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

correlated with follow up LVEF (r=0.33, p<0.001)(386).

In summary, this study demonstrates that CMR assessed longitudinal myocardial strain

predicts myocardial recovery. In addition, the feasibility of CMR assessment of longitudinal

strain established in this study may form the basis of further studies exploring the prognostic

role of longitudinal strain in DCM. Echocardiography assessed longitudinal strain, combined

with CMR late gadolinium enhancement (LGE), has been shown to predict outcome in

patients with dilated cardiomyopathy(383). LV longitudinal strain assessed by CMR feature

tracking has also been shown to be an independent predictor of survival in DCM(164).

However, DENSE has been shown to be more reproducible than feature tracking(387) and

further exploration of this is a promising area for future research.

4.5.2 Strengths of this study

One of the main strengths of this CMR study is the assessment of cardiac function, interstitial

fibrosis, replacement myocardial fibrosis, and myocardial strain in one study, enabling a

comprehensive, state of the art imaging evaluation of potential predictors of LV remodelling.

Another key strength is its prospective study design, with the inclusion of a DCM ‘inception

cohort’(388). This means that there is no survival bias that can occur in retrospective cohort

studies of LV remodelling, whereby only patients who survived to remodel have repeated

estimates of LV function. This therefore ensures that the estimate of the proportion of

patients who exhibited LV reverse remodelling is also not biased.

The number of patients who experienced improvement in LV function in this study was at the

296 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

higher end of estimates from previous studies, although still consistent with previous studies

(~60% of patients showed an absolute improvement of LVEF greater than 10 units). This

may reflect the composition of this cohort, with most patients in NYHA class I/II. This may

be viewed as a limitation (e.g. results not applicable to advanced DCM), but also a strength,

as highlighted in an editorial by Wilcox and Yancy, stating that ‘the future of successful

cardiac recovery programs includes all stages of HF, especially stage B or asymptomatic LV

dysfunction, when certain pathophysiological mechanisms may still be reversible’(388).

Another strength of the current study is the standardized baseline phenotyping using CMR

assessed LVEF contractile reserve, thereby reducing variability in LVEF assessment

compared to previous echocardiography based studies. CMR is the established gold standard

for assessment of cardiac volumes and function.

Further strength lies in the careful inclusion criteria whereby patients with a potentially

reversible cause of DCM were excluded. Previous studies of remodelling in DCM have

included cases where there is uncertainty regarding reversible disease causality, such as toxin

induced LV dysfunction, which may limit the generalizability of findings to all-cause DCM.

Many studies evaluating predictors of LV remodelling have done so in the context of

response to advanced heart failure therapy such as CRT or LVAD(389-393). Whilst such

studies make important contributions to the field as a whole, identifying that a favourable

response often occurs in females, without a history of ischaemic heart disease, with left

bundle branch block and lower LA volumes, they do not have direct applicability to patients

with ambulatory DCM.

297 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

In contrast to these prior studies, the focus of the current study has been to evaluate predictors

of LV reverse remodelling in DCM patients earlier in their disease course. This may have

identified predictors that the studies of advanced heart failure cannot. For example, it is as yet

unclear at what stage in the DCM disease course is recovery still possible. In addition,

patients in these earlier studies could often not undergo CMR due to device contraindications,

therefore the variables such as fibrosis in the current study would not have been evaluated.

Another important strength in this study is the statistical design and analysis, in which no

threshold for LVRR was defined. Previous studies have had variable definitions of LVRR,

making between study comparisons difficult. By not making such an assumption, this study

has evaluated predictors of overall LV functional improvement, thereby ensuring that the

results from this study can be readily compared to other studies. Whilst a threshold for LVRR

was not defined, the improvement in LV function remained clinically meaningful. All 9

patients with an initial LVEF <35% improved LV function to move out of current guideline

criteria for ICD implantation.

4.5.3 Limitations of this study

There are a number of potential limitations to this study which I will review in detail.

The first potential limitation relates to ascertainment bias. This could have arisen from

academic centre referral bias (recruitment via referrals to the CMR unit at a tertiary hospital

or from the Cardiomyopathy Service at a tertiary hospital). To address this, recruitment was

also performed via referral from Cardiology Consultants in a network of regional general

hospitals. In addition, a significant minority of patients self-referred via the Cardiomyopathy

298 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

UK website. The final DCM cohort consisted of patients across the mild-moderate-severe

disease severity spectrum, suggesting that there was not significant referral bias.

Another source of bias results from the inclusion criteria. As the primary hypothesis was to

evaluate LVEF contractile reserve, it was imperative that the most accurate assessment of

cardiac volumes and function was made with CMR. Therefore patients with atrial fibrillation

were excluded, due to variable RR intervals and the possibility of inaccurate quantification of

cardiac volumes. However, atrial fibrillation can be found in ~10% of patients with

DCM(394), therefore the generalisability of these findings to DCM patients with AF is not

known.

The second limitation arises from the method chosen to evaluate contractile reserve

(pharmacological dobutamine stress with evaluation of change in LVEF). Exercise stress has

been shown to be a more physiologically appropriate and demanding stress agent in the

setting of ischaemic heart disease(395) as it may invoke a more complete contractile

response. Whilst this would have been an interesting study design, as yet there are no

facilities for exercise-stress CMR in our CMR unit. Furthermore, exercise stress may be

submaximal, hampered by poor patient effort, limiting the utility of this technique.

In recognition of the limitations of load dependent LVEF contractile reserve

measurements(396), a novel method was developed to estimate strain contractile reserve,

which yielded insight into the importance of longitudinal strain in DCM. Reproducibility of

the strain measures will, to some extent, determine the diagnostic utility of DENSE derived

strain within this context, particularly given the relatively small changes in strain between

DCM and healthy cohorts.

299 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

With regards to dobutamine stress, it could also be argued that variability in contractile

reserve response could reflect insufficient stress, as opposed to true physiological impairment

of contractile reserve. However, all patients had a minimum 10 minutes of dobutamine

infusion and markers for response (heart rate or blood pressure change) were seen in all

participants. In this context, invasive haemodynamic cardiac catheter measurements of

contractile reserve would ideally have been included. This could form the basis of a future

study, though haemodynamic cardiac catheter based methods are unlikely to be of routine

clinical utility due to their inherent risks in a patient population that has no other indication

for them.

Another potential limitation is that the follow up assessment of LVEF with either 3d

echocardiography or CMR could introduce variability in the follow up LVEF measure.

However, this was assessed and the follow up imaging modality (CMR or echo) was not

associated with follow up LVEF on regression analysis. Whilst a limitation, this was also

unavoidable as even though all patients had MRI conditional devices, the safety of scanning

these devices at 3T had not been established.

The strengths of this study are also its limitations. The study was not designed to evaluate the

role of CRT in LV reverse remodelling in DCM or the utility of CMR contractile reserve to

predict response to CRT. Only 3 patients had CRT during follow up. CRT was therefore only

included in the analysis as a variable of interest. Similarly, patients with cardiac devices at

baseline were not included in the study, therefore introducing selection bias into the study

cohort. To address this, a separate study arm consisting of DCM patients with cardiac devices

could have been introduced, with 3d echocardiography assessment of contractile reserve. At

the time of study design, this was explored but not deemed feasible.

300 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

Given the largely asymptomatic/minimally symptomatic patient population with moderate

LV dysfunction, the findings of this study describe contractile reserve as a predictor of

remodelling in ambulant DCM patients. The findings are not necessarily applicable to an end-

stage DCM cohort.

Established and novel cardiac biomarkers were not evaluated as predictors of LV remodelling

in this study. This will be addressed in future work. However, it will be important to study a

broad range of biomarkers and interpret such results with caution, with careful adjustment for

potential confounders as it is unclear how much these biomarker signatures would reflect

current disease status as opposed to potential for recovery. For example, contractile reserve in

this study was not correlated with baseline LVEF, making it a marker that reflected

pathophysiology beyond LVEF alone. Lower levels of many cardiac biomarkers may simply

reflect less LV dysfunction, lower symptom class or shorter disease duration and not

independent pathology.

Due to the time course of the study, the study was also not designed to evaluate either the

prognostic role of contractile reserve in DCM, or whether the remodelling observed is

sustained recovery (cure) and associated with improved clinical outcomes or temporary

recovery on medical therapy (remission). There is emerging evidence to support the notion

that LV remodelling is associated with improved clinical outcomes, with reduced events

observed in a heart failure cohort with recovered ejection fraction(397) and DCM specific

cohorts with LVRR(398,399). However, this is not definitive, as CRT ‘super-responders’ do

also experience appropriate ICD shocks on follow up, suggesting that reverse remodelling

does not remove all clinical risk(400). In these scenarios, other non-LVEF factors such as

replacement myocardial fibrosis could underpin the persistent adverse outcome (for example

301 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

in the promotion of re-entrant arrhythmia).

This remains an important unanswered question which should form the basis of further

prospective study, though in an ambulatory DCM cohort such as this, the event rate is likely

to require multi-centre study design. A crucial unanswered question is whether the risk of

DCM ‘resets’ upon recovery of LV function, or whether impaired LVEF, at any stage,

portends ongoing increased risk of adverse events. Another remaining question is whether

myocardial recovery simply represents reversal of dysfunction or an additional process

(beyond repair/response to an initial insult). To address such a question, a pathophysiological

model of LV dysfunction that adequately captures the complexity of idiopathic DCM would

be required.

Finally, the results of this study are single centre, and although the study was appropriately

powered, the findings require replication in an independent cohort.

4.6 Summary

LV contractile reserve can be safely assessed by low-dose dobutamine stress CMR and is an

independent predictor of LV remodelling in patients with recent onset DCM. LV contractile

reserve assessed by longitudinal strain and relative RV contractile reserve are also predictive

of LV remodelling, but baseline myocardial strain, replacement myocardial fibrosis and

interstitial fibrosis are not predictive of follow up LVEF in this cohort. In this cohort, the LV

contractile reserve response is not related to baseline LVEF or the degree of interstitial

fibrosis measured on CMR.

302 Template by Friedman & Morgan 2014 Word Chapter 4: Imaging predictors of cardiac remodelling

4.7 Outline of further work

• Evaluate contractile reserve as a predictor of LV remodelling in the context of

established and novel biomarkers markers of cardiac disease.

• Validate the strain data from CMR cine DENSE against echocardiography strain

measures.

4.8 Acknowledgements

• Rick Wage, Chief CMR Radiographer, oversaw acquisition of all baseline and follow

up CMR studies.

• Consultant Cardiologists Dr Carl Shakespeare, Dr Amal Muthumala, Dr Jason Dungu,

and Dr Ravi Assomull actively recruited patients into the study.

• Research nurses Zohreh Farzad, Carmen Chan, Sally-Ann McRae, Iulia Muntenu and

Kevin Kirby assisted with participant preparation and were present during the

dobutamine scan as part of the safety protocol.

• Manivarmane Ramasamy performed all 3-d echocardiography studies.

• Peter Gatehouse, CMR Physicist, provided technical advice about the CMR

sequences, protocol and optimisations required in the context of dobutamine stress.

• Andrew Scott, CMR Physicist, developed the modified cine DENSE sequence as

outlined in the Methods.

303 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5 THE ROLE OF CLINICAL,

IMAGING, AND GENETIC DATA

IN PREDICTING CLINICAL

OUTCOMES IN DILATED

CARDIOMYOPATHY

304 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.1 Aims and hypotheses

The primary aim of this chapter is to describe the prognosis of titin cardiomyopathy in order

to inform patient stratification by genotype. The hypotheses are as outlined:

- Patients with DCM and titin truncating variants have an adverse event profile for the

composite endpoint of cardiovascular mortality, major heart failure and major arrhythmic

events, compared to patients with DCM without titin truncating variants.

- The effect on outcome of titin truncating variants in patients with DCM is modified in the

presence of additional environmental variables, including male gender, mid-wall

myocardial fibrosis, arrhythmia, and alcohol consumption.

5.2 Background

The preceding chapters have highlighted the importance of truncating variants in the titin

gene in dilated cardiomyopathy. It is the commonest genetic contributor to DCM. Therefore

in this chapter, I evaluate the prognostic implications of titin truncating variants in the context

of established clinical and imaging risk factors in DCM.

There is precedence for the use of genetic data in risk stratification in DCM, for example the

arrhythmic phenotypes associated with LMNA cardiomyopathy (49). As TTNtv are found in

up to 4 times as many patients with DCM compared to LMNA variants (14), risk stratification

using TTNtv would have broad implications in tailoring the clinical management of patients

with DCM.

305 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.3 Methods

5.3.1 Study cohort

For outcome analysis, the primary study cohort was a subset of the main study cohort (Figure

3-1), consisting only of patients recruited prior to 31st December 2014, to allow at least 1 year

of follow up time for all recruited subjects. Data on survival and events prior to 31st

December 2015 were recorded. Follow-up was censored at 31st December 2015 to allow for

data collection and adjudication of all events.

5.3.2 Variable definition and analysis

Variables from clinical, imaging, and genetic categories, defined in Chapters 2 and 3, were

evaluated in survival analysis (Table 5-1).

Table 5-1: Variables evaluated in survival analysis

Variables evaluated Methods Demographic/ Age, Gender, Ethnicity, NYHA class, a family history of As described in Clinical DCM, a family history of sudden cardiac death, a history of Chapter 3. sustained ventricular tachycardia, a history of atrial fibrillation, diabetes, hypertension, left bundle branch block, resting heart rate, beta blocker use, aldosterone antagonist use, diuretic use, ACE inhibitor use. CMR LVEF, RVEF, indexed left atrial volume, indexed LV end As described in diastolic volume, indexed LV end systolic volume, indexed Chapter 3. left ventricular mass, indexed left ventricular stroke volume, indexed RV end diastolic volume, indexed RV end systolic volume, indexed right ventricular stroke volume, mid-wall fibrosis late gadolinium enhancement, CMR features of left ventricular non-compaction

306 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Variables evaluated Methods Genetic Truncating variants in titin, lamin variants As described in Chapter 2. Blood Baseline serum creatinine Described in text.

Serum creatinine was measured using electrochemiluminescence immunoassay on a Cobas

8000 c702 modular analyser (Roche Diagnostics, Mannheim, Germany) following the

manufacturer’s protocol or determined from the point-of-care testing on the day of CMR

study.

5.3.3 Endpoints and adjudication of events

5.3.3.1 Endpoints

Endpoints were defined according to the 2014 American College of Cardiology/American

Heart Association definitions for cardiovascular endpoints in clinical trials(401) and the 2006

ACC/AHA/ESC guidelines for management of patients with ventricular arrhythmias (330).

They are defined in detail in the Appendices (p373). The primary endpoint was a composite

of cardiovascular mortality, major arrhythmic events and major heart failure events.

Major arrhythmic events were defined as:

o Haemodynamically stable and haemodynamically unstable ventricular

tachycardia

o Ventricular fibrillation

o Appropriate implantable cardiac defibrillator (ICD) discharge

o Aborted sudden cardiac death

307 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Major heart failure events were defined as:

o Heart transplantation

o Left ventricular assist device implantation

o Unplanned heart failure hospitalisation

Cardiovascular mortality and the arrhythmic and heart failure composites were predefined

secondary endpoints.

5.3.3.2 Follow up data collection and adjudication

Follow up data was collected and curated in 4 stages. Initially, medical data was collected

from primary care records, hospital care records, and patient questionnaire. The main sources

of data were primary and hospital care records. All primary care providers were contacted at

least once to request all patient medical records since the date of the patient’s recruitment to

the study. In the case of non-responding primary care practices, up to 2 further requests for

information were made. Secondary and tertiary care providers were contacted directly if the

primary care records received did not contain complete information about secondary care

admissions or consultations.

In addition, all patients were sent an annual questionnaire either by post or email (as per

patient preference) in which they were able to report any medical updates including hospital

admissions. The patient questionnaire served as a prompt to search for evidence of any events

not yet identified from the available medical records.

Survival status was also identified using the UK Health and Social Care Information Service

308 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

to ensure no deaths were missed. Death certificates and post mortem reports were obtained

where applicable.

This information was collated into a summary ‘Follow-up’ event form on the research

database by a team of up to 7 research nurses.

In the second stage, myself and 2 other research fellows reviewed all the Follow-Up event

data for each patient to confirm all major and minor events had been correctly recorded.

Events were reclassified in the event of inaccurate data entry. In addition, I reviewed a subset

(10%) of patient records with no events recorded to confirm that no major or minor events

were missed. In the third stage, I curated all primary end-point events to ensure each event

was linked to primary evidence of the event, as a final quality control step.

In the fourth stage, I presented the data to the independent data adjudication committee. A

committee of 3 cardiologists, each with expertise in electrophysiology, heart failure

management or clinical trial adjudication, blinded to imaging and genetic data, adjudicated all

primary end-point events. The adjudication committee established whether each arrhythmic

and heart failure event fulfilled pre-specified criteria using medical records including device

data where applicable, in line with the ACC/AHA guidance as outlined above(401) (330).

Cause of death was established from death certification, post mortem results, and medical

records, using the ACC/AHA guidance as outlined above(401) (330).

309 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.3.4 Statistical analysis

5.3.4.1 Survival analysis

Outcome analysis was conducted on the subset of patients who were recruited to the study

prior to 31st December 2014, to allow at least 1 year of follow up time for all recruited

subjects. For the survival analysis, event free survival was calculated from the date of study

entry to the date of the first event in the composite endpoint.

Data for all patients who were last known to be alive, or who had died after December 31st

2015, were censored on December 31st 2015.

The Kaplan–Meier method was used estimate the cumulative freedom from each endpoint

and the log-rank statistic was used to test the null hypothesis that there was no difference

between groups in the probability of an event at any time point. Kaplan Meier survival curves

were plotted and log rank tests performed using the survminer package in R.

The study design was a prospective observational study and the sample size obtained was the

maximal possible within constraints of resources and study timeframe to enable the

recruitment and assessment of patients with and without TTNtv.

5.3.4.2 Cox proportional hazard modelling

For the primary analysis, a Cox model was built to estimate the hazard ratios for each

endpoint associated with the presence of TTNtv.

310 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

To do this, an optimised baseline model was built using Cox proportional hazard modelling

evaluating clinical and demographic variables predicting the primary end point. This model

was built without inclusion of genetic data.

An outline of the steps to build the optimised baseline model predicting the primary endpoint

is shown in Figure 5-1. The same principle was applied to the secondary endpoints. All

modelling was performed using the survival package in R. The tmerge function was

used to create time dependent covariates.

There are various approaches to model selection. A purely automated variable selection

process was considered but rejected due to the limitations of this approach such as retention

of variables with inflated coefficients, model over-fitting, falsely narrow confidence intervals,

and not being able to incorporate scientific knowledge for important variables, data quality,

or collinearity between variables (402-405).

Model selection based on the purposeful selection of covariates, proposed by Hosmer and

Lemeshow, was the method chosen (Figure 5-1)(406,407). Whilst still based on stepwise

selection with the inherent limitations, the exact methods and criteria for variable inclusion

and exclusion have been defined so that the analysis is reproducible and objective. Genetic

data was then added to this baseline model.

The primary analysis and goal of this chapter was to evaluate the association of TTNtv with

the primary endpoint, on univariable and multivariable analysis. The secondary analysis was

to evaluate the association of TTNtv with the pre-specified secondary endpoints.

311 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Selec/on of variables of interest based on prior knowledge/literature

Univariable cox propor/onal hazard modelling of all variables of interest

All variables significant at p≤0.1 retained

Exclusion of correlated variables

Full model built with all significant variables

Purposeful backward stepwise selec/on with stepwise removal of least significant variable un/l only significant variables retained (p<0.05)

Model with only significant variables retained (p<0.05)

Addi/on of each previously discarded variable (p>0.1) to evaluate any effect on the coefficients of retained variables or the significance of the added variable in the presence of retained variables

Model tes/ng including tes/ng of propor/onal hazard assump/on (individual variables and overall model)

Transforma/on of any variable viola/ng propor/onal hazard assump/on

Tes/ng model with transformed variable(s)

Final op/mised model

Addi/on of variable of interest

Figure 5-1: Outline of variable selection for the baseline Cox proportional hazard model for analysis of

outcome endpoints.

312 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4 Results

5.4.1 Overview

5.4.1.1 Cohort description

In total, 633 patients with a confirmed diagnosis of DCM recruited prior to 31st December

2014 were included in the follow up analysis cohort. Of these, 8 patients were excluded from

analysis due to loss from follow up. In these patients, there was no follow up data despite at

least 2 requests for information from the primary care provider. This included patients who

had moved overseas. A further 21 patients were excluded from analysis because there was

missing data in any one of the endpoints. For example, we may have had information about a

heart failure hospitalisation, but we did not have complete medical records, therefore could

not be certain that the individual did not have another class of event such as an arrhythmic

episode. Prior to excluding these 29 patients from analysis, further efforts were made to

collect the data, including contacting the primary care providers and secondary care

cardiologists where such contact data was available. This resulted in 604 patients in the final

analysis.

5.4.1.2 Follow up time

In total, 604 patients were followed up for a median of 3.9 years (IQR 1.94 to 5.76 years).

Follow up time was truncated at 10 years given the reduced number of individuals with

follow up beyond 10 years (Figure 5-2). The decision to truncate at 10 years was made to

ensure that model assumptions were not violated due to sparse data but acknowledging the

compromise of losing events after the truncation date (n=2 patients who met the primary end

point).

313 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Figure 5-2: Histogram of follow up time in original dataset. Follow up time truncated at 10 years (dashed

red line).

5.4.1.3 Endpoints

Table 5-2: Table showing the number of patients meeting each endpoint (truncated dataset). Patients

were censored at the first event. *Cardiovascular death included 18 patients with heart failure death, 3

patients with sudden cardiac death, and 1 patient each with acute myocardial infarction, cerebrovascular

accident and cardiovascular –other.

Primary composite Cardiovascular death Arrhythmic Heart Failure

secondary secondary

Yes 78 24* 24 50

No 526 580 580 554

Overall, 78 patients met the primary composite endpoint (Table 5-2). Patients were censored

at first event, but may have had more than one event as outlined in Table 5-3.

314 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Table 5-3: Composition of events occurring during follow up leading to primary composite endpoint. Patients were censored at the first event.

CV death Heart Failure Cardiac LVAD Stable Unstable Aborted Appropriate Ventricular Count hospitalisation transplant sustained sustained sudden ICD activation fibrillation (n=78) VT VT cardiac death No Yes No No No No No No No 24 Yes No No No No No No No No 12 Yes Yes No No No No No No No 8 No No No No Yes No No No No 7 No No No No Yes No Yes Yes No 3 No No No No Yes Yes Yes Yes No 3 No Yes Yes No No No No No No 3 No No Yes No No No No No No 2 No Yes No Yes No No No No No 2 No No No No No No Yes Yes Yes 1 No No No No Yes Yes No No No 1 No No No No Yes Yes Yes Yes Yes 1 No No No Yes No No No No No 1 No No Yes Yes No No No No No 1 No Yes No No No Yes Yes Yes Yes 1 No Yes No No Yes No No No No 1 No Yes No Yes No Yes Yes Yes No 1 No Yes Yes No Yes Yes Yes Yes No 1 No Yes Yes Yes No No No No No 1 Yes No No Yes Yes Yes Yes Yes No 1 Yes Yes No No Yes No No No Yes 1 Yes Yes No No Yes No Yes Yes No 1 Yes Yes No Yes Yes No Yes No Yes 1

315 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.2 Building the baseline model to predict the primary end point

As outlined in Figure 5-1, the baseline model was built following evaluation of the

demographic, clinical and imaging variables in Table 5-1. The summary statistics of these

variables grouped by the primary end point are shown in Table 5-4.

The results of univariable regression Cox proportional hazard modelling of these variables for

the primary endpoint are shown in Table 5-5.

316 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Table 5-4: Baseline demographics in the outcome cohort, stratified by the primary endpoint. Data are

presented as mean (standard deviation) for continuous variables and number (percentage) for categorical

variables. For continuous variables, between group comparisons are made with the t-test or Mann

Whitney test. Categorical variables are compared with the Chi square test or the Fisher’s exact test if

there are fewer than 10 observations per cell. *Effect of serum creatinine evaluated separately in

sensitivity analyses due to high degree of missing data. SCD= sudden cardiac death, ACEi=ACE

inhibitor, LV/RV=left/right ventricle, EF= ejection fraction, EDVi=indexed end diastolic volume,

ESVi=indexed end systolic volume, SVi=indexed stroke volume, LVMi= indexed left ventricular mass,

LAVi= indexed left atrial volume, LVNC= left ventricular non compaction.

Variable All patients Primary Endpoint n=604 Yes (n=78) No (n=526) Mean (sd)/ Missing (%) Mean (sd)/ Mean (sd)/ P value n(%) n(%) n(%) Demographic Age (years) 53.45 (14.35) 0 54.9 (15.9) 53.2 (14.1) 0.29 Male gender 398 (65.9) 0 52 (66.7) 346 (65.8) 0.98

Caucasian ethnicity 524 (86.8) 0 66 (84.6) 458 (87.1) 0.68 Clinical Family history SCD 88 (14.6) 0 8 (10.3) 80 (15.2) 0.30 Family history 100 (16.6) 0 DCM 8 (10.3) 92 (17.5) 0.14 VT 11 (1.8) 0 6 (7.7) 5 (1.0) 0.001 NSVT 71 (11.8) 0 19 (24.4) 52 (9.9) <0.001 Atrial fibrillation 150 (24.8) 0 24 (30.8) 126 (24.0) 0.25 LBBB 164 (27.2) 0 21 (26.9) 143 (27.2) 1.00 Hypertension 180 (29.8) 0 30 (38.5) 150 (28.5) 0.097 Diabetes mellitus 71 (11.8) 0 12 (15.4) 59 (11.2) 0.38 Resting heart rate 74.55 (15.71) 33 (5.5) 72.6 (15.2) 74.8 (15.8) 0.30

NYHA class I 252 (41.7) 30 (5) 25 (32.1) 227 (43.2) II 230 (38.1) 28 (35.9) 202 (38.4) III 87 (14.4) 20 (25.6) 67 (12.7) 0.012

317 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Variable All patients Primary Endpoint n=604 Yes (n=78) No (n=526) Mean (sd)/ Missing (%) Mean (sd)/ Mean (sd)/ P value n(%) n(%) n(%) IV 5 (0.8) 2 (2.6) 3 (0.6) Medication Diuretic 270 (44.7) 0 47 (60.3) 223 (42.4) 0.005 Beta blocker 425 (70.4) 0 61 (78.2) 364 (69.2) 0.14 ACEi 478 (79.1) 0 68 (87.2) 410 (77.9) 0.09 Aldosterone 218 (36.1) 0 Antagonist 33 (42.3) 185 (35.2) 0.27 Imaging LVEF (%) 39.4 (12.4) 0 34.1 (13.1) 40.1 (12.1) 0.0001 LVEDVi (mL/m2) 127.3 (36.4) 0 143.3 (48.3) 124.9 (33.7) 0.003 LVESVi (mL /m2) 79.7 (36.5) 0 98.1 (47.3) 77.0 (33.8) 0.0002 LVSVi (mL /m2) 47.4 (13.0) 0 45.3 (14.4) 47.8 (12.8) 0.12 LVMi (g/m2) 90.8 (25.5) 14 (2.3) 99.1 (29.9) 89.6 (24.6) 0.007 RVEF (%) 52.0 (13.7) 13 (2.2) 47.0 (15.6) 52.7 (13.2) 0.002 RVEDVi (mL /m2) 87.5 (24.0) 13 (2.2) 90.4 (31.0) 87.0 (22.8) 0.64 RVESVi (mL /m2) 43.5 (21.6) 13 (2.2) 51.0 (29.7) 42.4 (19.9) 0.05 RVSVi (mL /m2) 44.0 (12.6) 13 (2.2) 39.6 (11.6) 44.7 (12.6) 0.001 Mid-wall fibrosis 213 (35.3) 0 LGE 42 (53.8) 171 (32.5) <0.001 LAVi (mL /m2) 60.9 (26.2) 24 (4.0) 76.9 (45.4) 58.6 (21.2) <0.00001 LVNC 28 (4.6) 0 3 (3.8) 25 (4.8) 0.95 Blood Creatinine 93.8 (29.6) 82 (13.6) 99.3 (30.5) 93.1 (29.4) 0.11 (µmol/L)*

318 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Table 5-5: Results of univariable Cox proportional hazard modeling of demographic, clinical, and

imaging variables predicting primary endpoint. HR= hazard ratio, CI= confidence interval

Variable Unit HR 95% CI P value

ACE inhibitor use Yes 1.74 0.9-3.38 0.10

Age per 10 years 1.09 0.93-1.28 0.3

Aldosterone antagonist use Yes 1.44 0.92-2.23 0.11

Atrial fibrillation Present 1.31 0.82-2.11 0.26

Beta blocker use Yes 1.39 0.82-2.34 0.22

Diabetes Mellitus Present 1.26 0.69-2.31 0.45

Diuretic Yes 2.1 1.34-3.29 0.001

Ethnicity Caucasian 0.79 0.43-1.45 0.44

Family history of DCM Present 0.47 0.23-0.99 0.05

Family history of sudden cardiac death Present 0.64 0.31-1.32 0.23

CMR features of left ventricular non compaction Present 0.89 0.28-2.81 0.84

Gender Male 0.98 0.62-1.56 0.93

Heart rate Per 1 bpm 0.99 0.98-1.01 0.40

History of hypertension Present 1.66 1.06-2.61 0.03

History of sustained ventricular tachycardia Present 4.7 2.04-10.85 <0.001

Indexed left ventricular end systolic volume per 10mL /m2 1.13 1.08-1.19 <0.0001

Indexed left atrial volume per 10mL /m2 1.11 1.06-1.15 <0.0001

Indexed left ventricular end diastolic volume per 10 mL /m2 1.11 1.06-1.16 <0.0001

Indexed left ventricular mass per 10g/m2 1.13 1.05-1.21 0.002

Indexed left ventricular stroke volume per 10mL /m2 0.86 0.73-1.01 0.07

Indexed right ventricular end diastolic volume per 10mL /m2 1.03 0.94-1.14 0.47

Indexed right ventricular end systolic volume per 10mL /m2 1.16 1.06-1.27 0.001

Indexed right ventricular stroke volume per 10mL /m2 0.73 0.61-0.86 <0.001

Left bundle branch block Present 1.08 0.66-1.76 0.77

Left ventricular ejection fraction per 10% 0.69 0.58-0.82 <0.0001

Mid wall fibrosis LGE Present 2.28 1.47-3.54 <0.0001

NYHA Class IV 7.93 1.32-47.66 0.02

319 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Variable Unit HR 95% CI P value

NYHA Class III 3.66 1.09-12.3 0.04

NYHA Class II 1.55 0.47-5.08 0.47

NYHA Class I 1.26 0.38-4.16 0.71

Right ventricular ejection fraction per 10% 0.76 0.65-0.89 <0.001

The variables that were significant at a threshold of p<0.10 were considered for inclusion in

full model building. At this stage, any variables with potential for inclusion in the next stage

of model building were reviewed for co-linearity. In addition, their appropriateness for

inclusion based on literature review was re-evaluated. Following this review, I decided to

exclude left and right stroke volume as they are highly correlated with left and right

ventricular ejection fraction, as well as being subject to heart rate variability and a greater

susceptibility to loading conditions compared to ejection fraction. Left ventricular end

diastolic and end systolic volumes are also correlated therefore only one of these was

retained. Either variable may have been appropriate and both represent left ventricular

dilatation. Left ventricular end systolic volume was chosen as previous studies have

demonstrated its prognostic importance(408,409). Left ventricular mass was excluded as it is

correlated with ventricular dilation. For right ventricular indices, right ventricular ejection

fraction was retained based on previous work from our group demonstrating its prognostic

importance over and above right ventricular volume indices(113).

Therefore the final variables included in the full model, prior to reverse stepwise selection

were left ventricular ejection fraction, right ventricular ejection fraction, mid wall fibrosis late

gadolinium enhancement, indexed left ventricular end systolic volume, indexed left atrial

volume, a family history of dilated cardiomyopathy, a history of sustained ventricular

tachycardia, NYHA class, a history of hypertension, diuretic use, and ACE inhibitor use.

320 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

From this, stepwise selection was performed, removing the least significant variable in turn

until only significant variables were remaining (p<0.05). This occurred in the order of

hypertension, ACE inhibitors use, diuretic use, family history of dilated cardiomyopathy,

indexed left ventricular end systolic volume, NYHA status, and right ventricular ejection

fraction.

This left the baseline model of left ventricular ejection fraction, indexed left atrial volume,

mid wall fibrosis late gadolinium enhancement and a history of sustained ventricular

tachycardia (Table 5-6).

Table 5-6: Adjusted baseline Cox proportional hazard model predicting the primary endpoint

Variable Unit HR 95% CI P

Left ventricular ejection fraction Per 10% 0.7 0.58-0.84 0.0002

Indexed left atrial volume Per 10mL/m2 1.11 1.06-1.17 0.00001

Mid wall fibrosis late-gadolinium enhancement Present 2.26 1.41-3.64 0.0008

Sustained ventricular tachycardia Present 3.24 1.28-8.24 0.01

Each previously discarded variable was added to the model in turn (hypertension, ACE

inhibitor use, diuretic use, family history of dilated cardiomyopathy, indexed left ventricular

end systolic volume, NYHA class, and right ventricular ejection fraction), as well as variables

that were not significant on univariable analysis. None of these variables became significant

on addition to the baseline model, nor did their inclusion affect the hazard ratios or p value

for the remaining variables (which may have suggested a possible interaction or confounder

effect). As left and right ventricular ejection fraction are correlated, the forced inclusion of

both variables as well as the inclusion of either variable was tested (Table 5-7). The model

321 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

with just left ventricular ejection fraction had a marginally lower AIC (Akaike information

criterion) compared to the model with just right ventricular ejection fraction. In the model

with both right and left ventricular ejection fraction, neither variable remained significant.

Therefore left ventricular ejection fraction only was retained, based partly on these results as

well as the overwhelming existing evidence for its utility as a prognostic marker in dilated

cardiomyopathy and heart failure.

Table 5-7: Comparison of Cox models with forced inclusion of left and right ventricular ejection fraction

(LV/RVEF). AIC= Akaike information criterion.

Baseline Model plus: Number of variables AIC

LVEF only 4 727.9822

RVEF only 4 728.5374

LVEF and RVEF 5 727.8523

LVESVi, LVEDVi and LVMi were not included in the multivariate model to avoid

colinearity, but separate models substituting these variables individually for LVEF showed

that LVESVi and LVEDVi were independently associated with the primary outcome, though

with smaller effect sizes and higher p values than LVEF (HR and 95% CI for primary

endpoint: LVESVi per 10mL/m2= 1.09, 1.03 to 1.16, p=0.001; LVEDVi per 10mL/m2= 1.07,

1.01 to 1.13, p=0.02). LVMi was not predictive of the primary outcome (HR and 95% CI;

LVMi per 10g/m2 = 1.04, 0.94-1.15, p=0.42).

At this stage, the baseline model was tested using the cox.zph() function in R to see if the

proportional hazard assumption was met. This function correlates the corresponding set of

scaled Schoenfeld residuals with time, to test for independence between residuals and time,

322 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

for each covariate and the global model. The output from this test was non-significant (Table

5-8), indicating the absence of evidence to contradict the proportionality assumption, either

for individual variables or the global model. This is illustrated graphically in Figure 5-3.

Table 5-8: Results of testing proportionality assumption of baseline model

rho chisq P value Indexed left atrial volume 0.0269 0.0423 0.84 Mid wall fibrosis late- gadolinium enhancement -0.1346 1.2337 0.27 Left ventricular ejection fraction 0.1033 0.8024 0.37 Sustained ventricular tachycardia 0.0563 0.2237 0.64 GLOBAL NA 2.2894 0.68

323 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Global Schoenfeld Test p: 0.6827

Schoenfeld Individual Test p: 0.837 Schoenfeld Individual Test p: 0.2667 4

0.8 2

0.4

Beta(t) for lavi10 Beta(t) for 0 Beta(t) for lgeYes Beta(t) for

0.0

−2 0.12 0.42 0.94 1.9 2.4 3.1 4.6 8.1 0.12 0.42 0.94 1.9 2.4 3.1 4.6 8.1 Time Time Schoenfeld Individual Test p: 0.3704 Schoenfeld Individual Test p: 0.6363

1 15

10 0

5 Beta(t) for vtYes Beta(t) for Beta(t) for lvef10 Beta(t) for −1

0

0.12 0.42 0.94 1.9 2.4 3.1 4.6 8.1 0.12 0.42 0.94 1.9 2.4 3.1 4.6 8.1 Time Time

Figure 5-3: Graphical diagnostic of the baseline Cox regression model predicting the primary endpoint.

The y axis represents the scaled Schoenfeld residuals, plotted against time on the x axis. The solid line is a

smoothing spline fit and the dashed lines are +/- 2 standard error bands. There are no systematic

departures from the fit line over time. ‘Lavi10’ and ‘lvef10’ are continuous variables, ‘lgeYes’ and

‘vtYes’ are categorical variables and their displays reflect this (2 strata).

In summary therefore the final baseline model predicting the primary endpoint consisted of

LVEF, mid wall fibrosis LGE, LAVi and sustained ventricular tachycardia. This model did

not violate proportional hazard assumptions. The variables in this baseline model suggest that

this is a representative DCM cohort as these predictors are in line with previously reported

predictors of outcome in DCM.

In the next section, I evaluate the prognostic importance of the presence of TTNtv, in

univariable analysis and when added to the optimized baseline model.

324 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.3 Analysis of the association between TTNtv and the primary end point

5.4.3.1 Titin and clinical outcomes: primary endpoint

Of the 604 patients in the outcome cohort, 71 patients had TTNtv (11.8%). Of these, 9

patients (12.7%) with TTNtv met the primary endpoint compared to 69 patients (12.9%)

without TTNtv. Survival curves comparing the freedom from the primary endpoint for

patients with and without TTNtv are shown in Figure 5-4.

Figure 5-4: Survival curves comparing freedom from the primary endpoint in patients with and without

TTNtv. Curves are compared using the log rank test. Confidence intervals are shown as dashed lines.

There is no significant difference between the two groups for the primary endpoint on unadjusted

analysis.

325 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

The 5 year event rate (95% confidence interval) for patients with TTNtv was 10.4% (2.0-

18.1%) compared to 13.6% for patients without TTNtv (10.1-16.9%).

On univariable cox proportional hazard modeling, the presence of a TTNtv was associated

with a hazard ratio of 0.81 (0.41 to 1.63, p=0.56) for the primary endpoint. On multivariable

analysis, adjusting for variables that predict the primary endpoint in this cohort (the baseline

Cox model), the presence of a TTNtv was not associated with the primary endpoint (HR 0.92,

0.45 to 1.87, p=0.82).

Therefore, in this cohort, the presence of a TTNtv alone was not a prognostic indicator. Given

the results of Chapter 3, outlining an interaction between titin and alcohol, and previous data

suggesting gender specific outcome effects in TTNtv DCM(79), exploratory analyses were

then performed to test the hypothesis that TTNtv confer an adverse prognosis in the presence

of an additional risk variable.

5.4.3.1.1 Titin, gender and the primary endpoint

It has previously been suggested that TTNtv are associated with adverse outcomes in males,

particularly for heart failure endpoints (79). Therefore the prognostic effect of TTNtv,

stratified by gender, was evaluated.

Of the 71 patients with TTNtv, 19 were female. Of these, only 1 female patient (5.3%) met

the primary endpoint, compared to 8/52 males (15.4%) with TTNtv. Cox proportional hazard

analysis showed no differences in primary outcome between these groups (p value for

TTN:gender interaction=0.36, Figure 5-9, Figure 5-5) though the comparison is limited by

326 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

the low event rate and number of females in the TTNtv cohort. On inspection of the survival

curves, males with TTNtv seem to have reduced freedom from the primary endpoint after 8

years of follow up but this is likely secondary to the small numbers of individuals with long

term follow up.

Figure 5-5: Survival curves showing freedom from the primary endpoint comparing patients with and

without TTNtv stratified by gender. Curves are compared using the log-rank test. The confidence

intervals of the curves have not been plotted for clarity but they overlap between all groups.

5.4.3.1.2 Titin, alcohol and the primary endpoint

Amongst patients with TTNtv, there was no evidence of a significant difference in outcome

when stratified by a history of alcohol excess (Figure 5-6). However, there were only 12

patients with TTNtv, a history of alcohol excess, and outcome data, and amongst these

patients, only 2 met the primary endpoint (17%), compared to 7 patients with TTNtv without

a history of alcohol excess (12%). There is therefore insufficient data to evaluate this

interaction further within this cohort.

327 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Figure 5-6: Survival curves showing freedom from primary endpoint in subset of patients with TTNtv,

stratified by a history of alcohol excess (EtOH) as defined in Chapter 3. Curves are compared using the

log rank test.

Of note, a history of alcohol excess in isolation was not associated with an adverse outcome

compared to patients without alcohol excess (Figure 5-7).

Figure 5-7: Survival curves showing outcome in DCM cohort stratified by a baseline history of alcohol

excess (EtOH). There is no significant difference between groups for the primary endpoint. Curves are

compared using the log rank test.

328 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.3.1.3 Titin, fibrosis and the primary endpoint

Mid-wall fibrosis LGE has been shown to be an important predictor of outcome in

DCM(147,264). In this cohort, this was replicated, with the presence of mid-wall fibrosis

LGE an independent predictor of adverse outcome (adjusted HR 2.26, 95% CI 1.41 to 3.64,

p=0.0008). There was no evidence however that the effect of TTNtv differed based on LGE

status (Figure 5-9, Figure 5-8).

Figure 5-8: Freedom from primary endpoint in DCM stratified by TTNtv and baseline history of mid

wall fibrosis LGE (LGE). The presence of LGE is clearly associated with adverse outcome, irrespective of

TTNtv status. There is no evidence of an interaction between TTNtv and LGE on event-free survival.

Curves are compared using the log-rank test.

5.4.3.1.4 Titin, arrhythmia and the primary endpoint

Of the 71 patients with TTNtv, 37 (52%) did not have a history of baseline arrhythmia (atrial

fibrillation, sustained ventricular tachycardia, non-sustained ventricular tachycardia prior to

329 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

study enrollment) compared to 34 patients (48%) with TTNtv with a history of arrhythmia.

On follow up, 7 TTNtv patients with a history of arrhythmia (9.9%) met the primary end-

point (endpoints: 5 heart failure, 1 arrhythmia, 1 cardiovascular death) compared to 2 TTNtv

patients without a history of arrhythmia (2.8%) (arrhythmia endpoint). In outcome analysis,

the presence of baseline arrhythmia was associated with adverse outcome independent of

TTNtv status (HR 1.91, p=0.007). There was no evidence that the effect of TTNtv differed in

the presence of a history of arrhythmia (TTNtv HR without history of arrhythmia 0.49, vs.

TTNtv HR with history of arrhythmia 0.85, p=0.51).

On analysis of arrhythmia subtype, 14 patients with TTNtv had baseline non-sustained

ventricular tachycardia prior to enrollment (NSVT). Of these, 5 (36%) met the primary

endpoint compared to 4 (7%) of TTNtv patients without NSVT. Across the whole cohort, the

presence of baseline NSVT alone was associated with an adverse outcome (unadjusted HR

1.84, p=0.04). In patients with TTNtv, the estimated effect of baseline NSVT was stronger

(HR 4.40; HR of NSVT plus NSVT:TTN interaction) but this difference did not attain

statistical significance in this small sample size (p=0.24) (Figure 5-9). There was no outcome

difference when stratified by a history of AF.

330 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Figure 5-9: Forest plot of Cox proportional hazard modelling of TTNtv interactions. Cox proportional

hazard modelling evaluating the interaction between TTNtv and gender, non-sustained ventricular

tachycardia (NSVT) or mid-wall late gadolinium enhancement (LGE) on the primary composite endpoint

of cardiovascular death, major heart failure events and major arrhythmic events. The hazard ratios of

the paired variables (not the interaction terms) represent the hazard ratios of each variable in isolation

(e.g. HR of TTNtv in patients without LGE, HR of LGE in patients without TTNtv).

5.4.4 Sensitivity analyses for the association between TTNtv and the

primary endpoint

5.4.4.1 Exclusion of individuals with LMNA variants

Currently, LMNA variants are the only clinically actionable DCM genetic variants therefore

sensitivity analysis was performed after exclusion of 8 patients with rare variants in LMNA,

331 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

of whom 4 met the primary end-point. The results from the unadjusted and adjusted HR for

TTNtv on the primary end-point were very similar (unadjusted HR 0.85, p=0.64; adjusted HR

0.92, p= 0.83).

5.4.4.2 Adjusting for device therapy

During the study follow up period, 160 patients underwent implantation of a cardiac device

(n=85 CRT-D, n=24 CRT-P, n=51 ICD). Of these, 131 were implanted for primary

prevention.

In total, 144 patients had devices implanted prior to censoring. Of these, 30 patients (20.8%)

met the primary endpoint compared to 114 patients who did not meet the primary endpoint

(79.2%).

To evaluate whether this could have affected the results for the association between TTNtv

and the primary endpoint, the Cox regression analysis was repeated with inclusion of cardiac

device therapy as a time dependent covariate.

On univariable analysis, the implantation of a cardiac device during study follow up was

associated with a hazard ratio of 3.4 (95% CI 2.14 to 5.40, p<0.0001) for the primary

endpoint. On multivariable analysis, adjustment of the baseline model inclusive of TTNtv for

the time dependent covariate of device implantation, did not affect the association between

TTNtv and the primary endpoint (TTNtv adjusted HR for primary endpoint 0.99, p=0.97;

Table 5-9).

332 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Table 5-9: Adjustment of Cox model for the time dependent covariate of cardiac device implantation does

not affect the association between TTNtv and the primary endpoint.

Variable Unit HR (95% CI) P value

TTNtv Present 0.99 (0.49-2.01) 0.97

Indexed left atrial Per 10mL/m2 1.13 (1.07-1.18) <0.00001 volume Mid wall fibrosis Present 2.26 (1.4-3.65) 0.0009 late-gadolinium enhancement Left ventricular Per 10% 0.74 (0.61-0.9) 0.0028 ejection fraction Sustained Present 2.51 (0.97-6.49) 0.057 ventricular tachycardia Cardiac device Present (implanted 2.25 (1.34-3.78) 0.002 during follow up)

The analysis was then restricted to cardiac-resynchronisation therapy only. In total, 101

patients had CRT-P or CRT-D implanted prior to censoring. Of these, 21 patients (20.8%)

met the primary endpoint compared to 80 patients who did not meet the primary endpoint

(79.2%).

On univariable analysis, the implantation of a cardiac-resynchronisation device during study

follow up was associated with a hazard ratio of 2.9 (95% CI 1.76 to 4.83, p<0.0001) for the

primary endpoint. On multivariable analysis, adjustment of the baseline model inclusive of

TTNtv for the time dependent covariate of cardiac-resynchronisation device implantation, did

not affect the association between TTNtv and the primary endpoint (TTNtv adjusted HR for

primary endpoint 1.05, 95% CI 0.91 to 2.14, p=0.90).

333 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.4.3 Exclusion of individuals with baseline VT

It may be argued that the presence of sustained VT at baseline already identifies a ‘high-risk’

patient and that further risk stratification is redundant. This strategy does not however

account for patients who may have presented acutely and not yet received optimal medical

therapy. Nevertheless, to address this potential limitation, the primary analysis was repeated

after exclusion of 11 patients with sustained VT at enrollment. In this cohort, 72 patients met

the primary endpoint (12.1%). There were 70 patients with TTNtv (11.8%), of whom 9

patients met the primary endpoint (12.9%).

This resulting baseline Cox model consisted of the same remaining variables as the original

model. On multivariable analysis in this subset, the presence of a TTNtv was not associated

with the primary endpoint (TTNtv adjusted HR for primary endpoint 0.94, 95% CI 0.46 to

1.92, p=0.87) (Table 5-10).

Table 5-10: Sensitivity analysis: VT. Unadjusted and adjusted Cox proportional hazard model predicting

the primary endpoint in subset of cohort after exclusion of patients with baseline sustained VT at

enrollment. HR=hazard ratio; LVEF=left ventricular ejection fraction; LAVi=indexed left atrial volume;

LGE=mid wall fibrosis late gadolinium enhancement; TTNtv=truncating variant in titin gene.

Unadjusted Adjusted

Variable Unit HR 95% CI P HR 95% CI P

LVEF Per 10% 0.67 0.55 to 0.80 <0.0001 0.69 0.57 to 0.84 0.0002

LAVi Per 10mL/m2 1.11 1.07 to 1.16 <0.0001 1.12 1.06 to 1.17 <0.0001

LGE Present 2.26 1.42 to 3.59 0.0006 2.25 1.39 to 3.67 0.001

TTNtv Present 0.94 0.47 to 1.89 0.86 0.94 0.46 to 1.92 0.87

334 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.4.4 Adjusting for renal function

In 416 patients, serum creatinine was measured using electrochemiluminescence

immunoassay and in 106 patients, serum creatinine was determined from the point-of-care

testing on the day of CMR study. In 82 patients, serum creatinine data was unavailable. Due

to the high degree of missingness, serum creatinine was therefore not evaluated in the

primary models. Survival analysis was then repeated on the subset of data with complete

creatinine data.

In this cohort of 522 patients, 65 patients (12.5%) met the primary endpoint. There were 57

patients with TTNtv (10.9%), of whom 5 met the primary endpoint (8.8%). On univariable

analysis, serum creatinine was associated with the primary endpoint (HR per 10µmol/L

increase in serum creatinine: 1.09, 95% CI 1.00 to 1.18, p=0.04). On multivariable analysis,

adjusting for serum creatinine in the baseline Cox model, there remained no association

between TTNtv and the primary endpoint (Table 5-11). On multivariable analysis, serum

creatinine was not an independent predictor of the primary endpoint (Table 5-11).

Table 5-11: Association between TTNtv and the primary endpoint, adjusted for the baseline Cox model in

the subset of patients with complete serum creatinine data.

Variable Unit HR 95% CI P Left ventricular ejection fraction Per 10% 0.69 0.56 to 0.86 0.0007 Indexed left atrial volume Per 10mL/m2 1.11 1.07 to 1.17 <0.0001 Mid wall fibrosis late-gadolinium enhancement Present 2.45 1.43 to 4.20 0.001 Sustained ventricular tachycardia Present 3.41 1.21 to 9.62 0.02 Creatinine per 10µmol/L 1.02 0.93 to 1.12 0.62 TTNtv Present 0.67 0.27 to 1.69 0.39

335 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.5 Non-titin genetic variants and clinical outcomes: primary endpoint

The primary focus of the analyses in this Chapter was to evaluate the prognostic importance

of TTNtv. Other DCM genetic variants were identified (as outlined in Chapters 2 and 3) and

their association with the primary endpoint is shown for completion. However, further

interpretation based on this analysis is limited by the relatively smaller proportion of patients

with non TTNtv genetic variants.

Table 5-12: Numbers of patients with rare variants (MAF <0.0001) in non-titin DCM genes meeting the

primary endpoint. Data shown as count (percentages). Variant class: ms=missense variants, tv=

truncating variants, if no qualifier then variants reported include both variant classes.

Gene Primary endpoint = No Primary endpoint = Yes N=527 N=78 LMNA 4 (0.8) 4 (5.1) MYH7ms 21 (4.0) 1 (1.3) TNNT2ms 7 (1.3) 1 (1.3) DSPtv 5 (0.9) 0 (0.0) BAG3 4 (0.8) 1 (1.3) RBM20 17 (3.2) 4 (5.1)

On univariable and multivariable analysis, adjusting for the baseline Cox model predicting

the primary endpoint, only rare variants in LMNA were associated with the primary endpoint

(Table 5-13). The survival curve for individuals with and without LMNA variants is shown in

Figure 5-10. Although the confidence intervals for patients with LMNA variants are wide,

reflecting the small sample size, there is evidence of early progression to adverse events in

this subgroup. This reflects the large effect size with LMNA variants (adjusted HR for

primary endpoint 3.80). This is consistent with published reports on LMNA cardiomyopathy.

336 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Table 5-13: Association between non-TTNtv genetic variants and the primary endpoint. The adjusted

analysis is adjusted for the variables in the baseline Cox model (LVEF, LAVi, VT, LGE).

Unadjusted Adjusted

Gene HR 95% CI P value HR 95% CI P value

LMNA 3.80 1.39 to 10.41 0.009 3.80 1.04 to 13.92 0.04

MYH7 0.59 0.14 to 2.40 0.46 1.04 0.25 to 4.32 0.96

TNNT2 0.66 0.09 to 4.77 0.68 0.51 0.07 to 3.86 0.52

BAG3 2.22 0.31 to 15.98 0.43 1.09 0.15 to 8.07 0.93

RBM20 2.17 0.79 to 5.95 0.13 2.21 0.69 to 7.08 0.18

Figure 5-10: Survival curve for patients meeting the primary endpoint stratified by LMNA variant status.

Curves are compared using the log rank test.

337 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.4.6 Analysis of the association between TTNtv and secondary outcome

end points

In total, 24 patients met the secondary cardiovascular mortality endpoint, including 2 patients

with TTNtv (2.8%) and 22 patients without TTNtv (4.1%). Fifty patients met the major heart

failure secondary endpoint, including 6 patients with TTNtv (8.5%) and 44 patients without

TTNtv (8.3%). In total, 24 patients met the secondary major arrhythmic endpoint, comprising

3 patients with TTNtv (4.2%) and 21 patients without TTNtv (3.9%). There was no

significant difference between TTNtv positive and negative DCM patients for the secondary

endpoints of cardiovascular mortality, major heart failure events or major arrhythmic events

(Figure 5-11).

338 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

page 1 of 1 Outcome in DCM stratified by titin status: HF composite

Strata TTNtv negative TTNtv positive

1.00

0.75

0.50

0.25 p = 0.82 Freedom from HF endpoint

0.00 0 2 4 6 8 10 Years Outcome in DCM stratified by titin status: Arrhythmia composite

Strata TTNtv negative TTNtv positive

1.00

0.75

0.50

0.25 p = 0.82 Freedom from Arrhythmia endpoint Freedom from Arrhythmia 0.00 0 2 4 6 8 10 Years Outcome in DCM stratified by titin status: CVS mortality

Strata TTNtv negative TTNtv positive

1.00

0.75

0.50

0.25 p = 0.39 Freedom from CVS death

0.00 0 2 4 6 8 10 Years

Figure 5-11: Freedom from secondary endpoints stratified by TTNtv status. There is no difference

between TTNtv positive and negative groups for cardiovascular mortality or the pre-specified arrhythmic

and heart failure endpoints. Curves are compared using the log rank test.

339 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

5.5 Discussion

In this outcome study of 604 patients with DCM, we leveraged the power of informed variant

curation, detailed cardiac phenotyping with CMR, and careful adjudication of clinical

outcomes to understand the prognosis of TTNtv cardiomyopathy. As the commonest genetic

cause of DCM(5,79,84), improved understanding of the clinical manifestations of TTNtv

cardiomyopathy offers scope for refined patient stratification by genotype. We demonstrate

that overall the presence of TTNtv DCM has a similar event profile to non TTNtv DCM, in a

cohort of ambulant DCM patients with medium term (~4 year) follow up.

In addition, this study also highlights a marked improvement in DCM outcomes overall when

compared to previous registry and cohort data. This data may inform clinical management

and guide future studies of risk stratification in DCM to focus on reducing the burden of heart

failure complications.

5.5.1.1.1 Clinical outcomes in titin cardiomyopathy

The primary analysis in this chapter was to evaluate the incidence of major adverse cardiac

events in titin cardiomyopathy. Of the 604 patients in the outcome cohort, 71 patients had

TTNtv. There was no difference in the number of patients with TTNtv who met the primary

composite endpoint (cardiovascular mortality, major arrhythmic events, major heart failure

events) compared to patients without TTNtv.

This suggests therefore that patients with TTNtv DCM have a similar prognosis to patients

without TTNtv DCM. In this cohort, the presence of a family history of DCM was associated

with the primary endpoint on univariable analysis, suggesting that there may be a genetic

340 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

influence on prognosis in DCM, though the association was not robust to multivariable

analysis.

With the exception of lamin cardiomyopathy, no other genetic variants in DCM have been

reported to be associated with an adverse prognosis. In this cohort, non TTNtv genetic

variants were not the primary focus of analysis, though we were also able to replicate the

adverse prognosis of LMNA cardiomyopathy, despite the very small numbers, a reflection of

the large effect size. We did not demonstrate an adverse prognosis for any other DCM genetic

variant.

The absence of association between TTNtv and an adverse prognosis is in line with the

phenotype data described in Chapter 3, in which the phenotype of TTNtv DCM was not more

severe than non-TTNtv DCM, except in the presence of an important environmental modifier.

Therefore exploratory analysis was performed to test the hypothesis that the prognosis of

TTNtv DCM was altered in the presence of an environmental modifier. It must however be

stressed that conclusive analysis of the role of environmental modifiers was not possible with

the sample and event size in this study.

Titin and gender

Our cohort of TTNtv DCM patients were 71% male. The gender distribution is interesting

because DCM is known to have a 3:1 male:female ratio, yet one might expect an autosomal

dominant inheritance genetic form of DCM to have a 1:1 male to female ratio. This suggests

that additional factors may be important for the development of phenotype in individuals with

TTNtv and the presence of TTNtv alone may not be sufficient. It is possible that female

341 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

gender confers a protective effect in the presence of TTNtv, or that males with TTNtv are

susceptible to a greater extent to an additional genetic modifier and/or have greater

exposure/susceptibility to an environmental modifier. Previous studies have reported

conflicting outcomes for TTNtv DCM stratified by gender(79,84-86). There may be an

important difference in outcome profile between males and females with TTNtv suggested by

the 10% difference in event rate observed in this cohort. However, in this largest single centre

study of TTNtv DCM, with long follow up and carefully adjudicated endpoints, the event rate

was not sufficient to show a gender protective effect on outcomes in TTNtv DCM patients.

Titin and mid-wall fibrosis

There was no evidence of difference in the presence of left ventricular mid wall fibrosis, a

marker of adverse outcome in DCM(13), between TTNtv positive and negative groups. In

addition, there was no significant interaction between adverse event rates in TTNtv patients

with and without mid-wall fibrosis LGE. Importantly, this suggests that LGE remains an

independent predictor of outcome in DCM, regardless of TTNtv status. This data is critical as

we begin to understand the multi-modality nature of risk stratification in DCM(347).

Titin and alcohol

Whilst there is evidence of a phenotypic interaction between TTNtv and alcohol (Chapter 3),

the sample size of TTNtv patients with a history of alcohol excess did not permit definitive

analysis of the prognostic implications of a TTNtv:alcohol interaction. This remains an area

of interest and may be addressed in larger, multicentre studies of outcome in TTNtv DCM.

342 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Titin and arrhythmia

An important aspect of the clinical management of DCM is the early identification of the at

risk patient. In this outcome study, we find that TTNtv DCM is not associated with an

increased risk of long term arrhythmic events (secondary endpoint) compared to patients

without TTNtv DCM.

In this cohort, exploring the interaction between TTNtv and arrhythmia further, amongst

TTNtv patients with early NSVT, there was an increased rate of adverse events (primary

endpoint) compared to patients without TTNtv with early NSVT, though this did not attain

statistical significance. This suggests there may be a subtle interaction between TTNtv and

arrhythmia generation requiring further exploration in larger studies with longer follow up.

The absence of a significant association between TTNtv and arrhythmic endpoints may

reflect that either TTNtv are not associated with long term arrhythmic risk, or this finding

may be driven by the overall low arrhythmic risk in this cohort, explored further in the

limitations section. Future studies focusing on subpopulations of DCM with a higher risk of

arrhythmia are therefore required to elucidate a potential interaction more fully. For example,

TTNtv may have a role to play in risk stratification within a higher risk subset of DCM

patients, such as those with low LVEF and mid-wall fibrosis.

5.5.1.1.2 Overall adverse event rate in DCM cohort and predictors of outcome

The overall baseline Cox model to predict the composite endpoint in this cohort consisted of

left ventricular ejection fraction, mid-wall fibrosis late gadolinium enhancement, left atrial

volume and a history of sustained ventricular tachycardia. These predictors are in line with

343 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

previously reported predictors of outcome in DCM, not just from our institution, and

demonstrate that this is a representative DCM cohort(13,127,147,410). In particular, left atrial

volume was the variable in the baseline model that had the lowest p value. This finding is

concordant with recent studies highlighting the importance of the left atrium as an emerging

cardiovascular biomarker(411,412).

Somewhat surprisingly, we find that the rate of adverse events in this cohort (13% of patients

met the primary endpoint, median follow up 3.9 years), particularly with regards to sudden

cardiac death (3 patients experienced sudden cardiac death), is much lower than previous

registry data(235), but also historic experience from our own institution(13,146), suggesting

that our findings are not simply a reflection of CMR bias.

To some extent, this is likely to reflect the milder, less symptomatic cohort of patients

recruited (the majority in NYHA I/II), but is also a reflection of optimal medical therapy,

with over 70% of patients on beta blocker therapy and almost 80% on ACE inhibitor

treatment. This improvement in outcome benefits patient care and highlights that future

studies of risk in this field should focus on reducing the increasing burden of heart failure

hospitalisations and heart failure deaths. This reduction in sudden cardiac death is also

mirrored in a recent study of over 40,000 heart failure with reduced ejection fraction patients

across 12 clinical trials between 1995-2014, demonstrating a reduction in the cumulative

incidence of sudden cardiac death from 2.4% in the earliest trial to 1.0% in the recent

trial(413).

The low arrhythmic event rate does have implications for the findings of our study, meaning

that we could not conclusively evaluate genetic and non genetic predictors of arrhythmia,

344 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

particularly the association between TTNtv and arrhythmia. This is despite having a

relatively broad inclusion criteria for the arrhythmic endpoint, including major arrhythmias

and aborted sudden cardiac death and not just sudden cardiac death. This is explored further

in the limitations section.

5.5.2 Strengths of this study

This study, commenced in 2009, was an enormous endeavour in terms of cost, resources,

physician and nursing staffing and of course patient contribution. To date, this is the largest

integrated study of clinical, imaging, and genetic predictors of outcome in a precisely

phenotyped cohort of patients with DCM with independently adjudicated event data.

5.5.3 Limitations of this study

There are a number of potential limitations to this work. The low arrhythmic event rate in this

cohort has been established. Whilst this may reflect a CMR bias, with exclusion of patients

from other institutions who may not have had a CMR prior to ICD implantation (in our

institution all patients have a CMR prior to ICD implantation where clinically possible), it

could be argued that optimal design of an arrhythmic risk analysis study should probably

exclude individuals who have already declared as arrhythmic risk. It is also possible that

previous estimates of risk in DCM were conflated by inadvertent inclusion of patients with

ischaemic heart disease into echocardiography based studies, which has been avoided through

careful CMR based phenotyping in this cohort. We have previously demonstrated that CMR

phenotyping identifies ~15% of patients with DCM diagnosed on another imaging modality

initially who have evidence of ischaemic heart disease despite a history of apparently

unobstructed coronary arteries(108).

345 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

The low event rate justifies our selection of a composite endpoint. The composite end point

was originally chosen based on previous studies from our institution, with an expected 20%

event rate at 5 years for the composite endpoint. However, the use of a composite end point

means that the findings of our study are not particularly informative for the prediction of

arrhythmic events, a major target for risk stratification in DCM.

The low event rate may also be a reflection of cardiac device therapy implanted during the

follow up period of the study, though on analysis with inclusion of device therapy as a time

dependent covariate, TTNtv were still not associated with the primary endpoint. Whilst

device therapy was adjusted for, there are other changes in therapy that may have occurred

during the course of the study that were not captured and therefore not adjusted for. For

example, temporal changes not only in medication dose during the course of a patient’s

treatment, but changes in medication class, for example the introduction of novel heart failure

therapy such as Ivrabadine or Entresto. The data collected could have spanned a 6 year time

frame (2009-2015) yet minimum follow up was 1 year. Therefore there is a possible 5 year

difference between the first and last patient, during which time changes in treatment could

have affected our findings.

Also, as with any model built using data from one time point, the model does not account for

time varying changes in event risk. Although genetic data is the exception to this, most other

variables evaluated are likely to vary during the time course of the study, which may affect

the risk of events, particularly sudden cardiac death. To avoid this would require resampling

of data at pre-specified intervals during the study, but this was not feasible during this study

and this method would still not fully account for all time varying changes in risk.

346 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

Importantly, a key limitation of this work is that this is a single centre study. All findings

require replication in an independent cohort. However, at present, this is the largest published

cohort of patients with DCM who have undergone CMR, genetic sequencing, and adjudicated

long term follow up and as such, a comparable replication cohort has not yet been identified.

The study design is a prospective observational cohort study. The sample size obtained was

in line with the target mega cohort sample size of n=1000 patients with DCM (n=928 patients

with confirmed imaging diagnosis of DCM recruited), which was the estimated number of

patients that could be recruited within the timeframe and resources of the study. Such a study

design enables us to draw inferences about the differences between patients with and without

TTNtv, as well as identify other associations from careful analysis of the rich dataset. Based

on our current sample size of patients with TTNtv, we are powered at 80% and an alpha level

of 0.05 to detect a hazard ratio for the association between TTNtv and the primary composite

endpoint of ≥2.15 (using the powerSurvEpi package in R). This hazard ratio would be

considered relatively low for risk stratification in DCM patients in a clinical context(250).

Therefore we can state that we are adequately powered within this study to detect a clinically

meaningful difference in outcomes should one exist between DCM patients with and without

TTNtv. A perceived limitation may be that the study had not been formally powered to detect

a difference in outcome between patients with and without TTNtv. However, at the time of

study design, there was no pilot data to support such a calculation. Going forwards, the

results of this study could help plan larger multi-centre studies evaluating TTNtv

cardiomyopathy. A sample size of at least 1826 patients with TTNtv would be required to

detect a difference based on the event rate observed in this cohort (using the

powerSurvEpi package in R), which could only be achieved through meta-analysis of data

from several projects.

347 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

These are challenges in observational studies, but they do not diminish the contribution that

the findings of this study make to the advancement of knowledge regarding titin

cardiomyopathy and risk stratification in DCM.

Other statistical challenges when analysing data from observational studies include missing

data, heterogeneity in samples and selection bias. These limitations with respect to

ascertainment and selection bias have been discussed in Chapter 3. Whilst the findings of

observational studies like this one may not be as accurate as those from clinical trials, they do

have greater external validity compared to those studies, meaning that the findings are

potentially more generalizable to clinical DCM cohorts.

Every effort was made during this study to evaluate key variables known to be implicated in

risk in DCM. However, a major exception is the absence of prognostic biomarker data

including natriuretic peptides and troponin. Subsequent to this analysis, this data has been

collected on a subset of the cohort (n=426 patients) and will form the basis of future work on

multi-modality risk stratification in DCM, incorporating biomarker, imaging, genetic and

clinical data. The analyses outlined in this Chapter were not performed in this biomarker

subset cohort to avoid model-overfitting. The number of primary endpoint events in the

cohort of n=426 was 44, meaning that only 4 variables could be evaluated to keep the

‘number of events per variable’ above the recommended threshold of 10(414). In the full

study cohort however, evaluating the prognostic importance of TTNtv in DCM, the number

of events per variable was 19, which mitigates against model overfitting.

Another major exception was ECG markers such as signal averaged electrograms,

fragmented QRS, QRS-T angle and T wave alternans data which were not recorded in the

348 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

cohort. Whilst they have been previously linked to adverse events in DCM, the individual

hazard ratios for these variables have been low(250), and certainly much lower than CMR

parameters, they have not been consistently replicated and are not widely available in clinical

practice. Therefore their absence from this dataset is not a major limitation in the wider

context of DCM risk research.

5.6 Summary

In this outcome study of 604 patients with DCM, we leveraged the power of informed variant

curation, detailed cardiac phenotyping with CMR, and careful adjudication of clinical

outcomes to understand the prognosis of TTNtv cardiomyopathy. While TTNtv are common

in DCM and may be used to confirm a genetic cause for DCM and for cascade screening they

do not appear to predict clinical outcomes in ambulant DCM cases with medium term (~4

year) follow up.

5.7 Outline of further work

• To harness the prognostic potential of the rich dataset, more complex machine

learning methods will be used to develop fuller models to predict risk in this cohort.

This will be performed in collaboration with researchers at the University of

California, San Francsico and the University of Oxford.

• Develop an integrated multi-modality risk score in DCM incorporating biomarker

data.

5.8 Acknowledgements

• The NIHR Royal Brompton Cardiovascular Biomedical Research Unit (BRU)

349 Template by Friedman & Morgan 2014 Word Chapter 5: The role of clinical, imaging, and genetic data in predicting clinical outcomes in dilated cardiomyopathy

research nurses recruited all patients to this study and together with BRU admin

support staff coordinated the retrieval and first stage data entry of follow up event

data from primary care, hospital records, coroners’ offices and national registry data.

• Clinical research fellows Dr Brian Halliday and Dr Amrit Lota contributed to the

initial quality control of the follow up data collected by the research nurses.

• Dr Julian Jarman, Dr Resham Baruah and Prof Michael Frenneaux sat on the

independent data adjudication committee, reviewing all endpoints.

• Steve Collins and Yasin Karafil developed and maintained the BRU database for

collation, storage and processing of all event data.

• Statistical mentorship and advice was provided by Simon Newsome.

350 Template by Friedman & Morgan 2014 Word Chapter 6: Summary: What this thesis adds to the field and future work

6 SUMMARY: WHAT THIS THESIS

ADDS TO THE FIELD AND

FUTURE WORK

The overall aim of this thesis was to evaluate whether the integration of clinical, genetic and

advanced imaging data could improve our understanding of DCM pathobiology, inform

patient stratification, and identify predictors of myocardial remodelling or clinical outcomes

in DCM.

Genetic architecture of DCM

To address these aims, the first part of this thesis defined the genetic architecture of DCM, by

evaluating the burden of rare genetic variation in putative DCM genes over and above

background population genetic variation. To date, this is the largest cohort study of DCM

patients to undergo standardised sequencing for almost 60 putative DCM genes. The study

confirmed that titin was the largest single gene contributor to DCM, a finding which

underpinned the subsequent phenotype study in this thesis. The study also provided an

updated frequency estimate of truncating variants in titin in an ambulant DCM cohort, as

opposed to the end-stage cohorts of the original reports.

In addition, the study confirmed rare variation in a further 4 genes with robust association to

351 Template by Friedman & Morgan 2014 Word Chapter 6: Summary: What this thesis adds to the field and future work

DCM and suggested that the remaining putative DCM genes require further evidence to

support their association with DCM. This is not to say that the remaining genes are not valid

DCM genes, but that in light of the degree of tolerated background population genetic

variation we now know to exist, a more critical evaluation of their contribution to DCM is

required. As extensively discussed, there were of course limitations to the analysis strategy

used. Going forwards, some of these will be addressed through planned studies using whole

genome sequencing (gnomAD) data as a comparator instead of whole exome sequencing

(ExAC) data, and the identification of replication cohorts from collaborating national and

international centres.

Together, these findings will be informative for clinicians trying to make sense of genetic

variation in their patients, for diagnostic sequencing laboratories deciding which genes to

prioritise for analysis, and for researchers designing studies in DCM.

Phenotype of titin cardiomyopathy

The thesis then moved on to detailed phenotype study of titin cardiomyopathy in over 700

DCM patients. As the commonest genetic contributor to DCM, the findings of this study will

be of great interest to clinicians who manage DCM patients. The novelty of the study lay in

the use of cardiovascular magnetic resonance for phenotype analysis, providing in vivo tissue

characterization and accurate quantification of cardiac chamber dimensions and function.

This enabled the discovery of a subtle but potentially important gene-environmental

interaction between titin and moderate excess alcohol consumption, whereby DCM patients

with titin truncating variants had more severely impaired left ventricular function in the

presence of alcohol excess, compared to patients with titin truncations alone or alcohol excess

alone. This was the first study to demonstrate a titin-alcohol interaction and these data add to

352 Template by Friedman & Morgan 2014 Word Chapter 6: Summary: What this thesis adds to the field and future work

the growing body of clinical and pre-clinical literature supporting a ‘two-hit’ hypothesis for

some cases of titin cardiomyopathy, whereby an additional environmental modifier is needed

for the development of the phenotype in the presence of a titin truncating variant(88,92).

Going forwards, the findings require replication in an independent DCM cohort, although our

research group have recently completed sequencing of an independent cohort of patients with

alcoholic cardiomyopathy and found a similar prevalence of titin truncating variants as in

DCM. It could be argued that this provides biological replication at least for a titin-alcohol

interaction. These findings provide clinically informative data to guide lifestyle advice for

affected patients and has guided further translational study, with a study planned to assess the

effect of alcohol in a rodent model of titin cardiomyopathy (collaboration with a Spanish

group).

The study also found that titin cardiomyopathy was associated with a limited hypertrophic

response meaning that affected patients had lower left ventricular mass and thinner left

ventricular walls, after adjustment for potential confounders. This finding supports recent

pre-clinical data(88) and advances our understanding of the pathobiology of titin

cardiomyopathy.

Imaging predictors of remodelling in DCM

The thesis then moved to evaluate remodelling in DCM, specifically imaging predictors of

left ventricular functional improvement at 1 year. This study demonstrated that low-dose

dobutamine stress was a safe and feasible method to evaluate contractile reserve in DCM

patients and that contractile reserve assessed this way was an independent predictor of

myocardial remodelling. This was the first CMR study to evaluate dobutamine stress-

contractile reserve. Unlike previous studies, it was therefore possible to evaluate interstitial

353 Template by Friedman & Morgan 2014 Word Chapter 6: Summary: What this thesis adds to the field and future work

and replacement fibrosis as remodelling predictors as well as perform in vivo tissue

characterisation to explore the biological basis of contractile reserve. In contrast to previous

research, in this cohort, myocardial fibrosis did not underpin contractile reserve, nor did it

predict remodelling. This study therefore added to the field by highlighting that predictors of

remodelling are distinct compared to predictors of outcomes in DCM, and added to our

understanding of the pathobiology of contractile reserve. The study also demonstrated the

feasibility of a novel cine DENSE CMR sequence to assess myocardial strain and showed the

importance of longitudinal strain in defining subtle abnormalities in DCM patients compared

to healthy volunteers, as well as as a useful predictor of response to therapy in DCM patients.

This sequence is now being used by researchers studying other cardiomyopathies. These

results advance our understanding of the pathobiology of DCM and may have utility in

detection of subclinical disease, particularly in genotype positive, phenotype negative

individuals.

Predictors of clinical outcomes in DCM

The thesis concluded with a study of the prognosis of titin cardiomyopathy in DCM. The

study demonstrated that there was no difference in the rate of adverse cardiovascular events

between DCM patients with and without titin cardiomyopathy. As the largest and most

comprehensive genotype-phenotype study of titin cardiomyopathy to date, these data provide

useful information to guide patient stratification, though it must be stressed that the overall

cohort was representative of a largely ambulant DCM cohort with medium term follow up

only. The number of events within this cohort was not sufficient to show an effect of TTNtv

on outcomes in DCM patients. Larger, likely multi-centre studies with longer term follow up

are needed to explore the prognostic role of potential titin-environmental interactions such as

alcohol consumption and to determine the long term prognosis of titin cardiomyopathy,

354 Template by Friedman & Morgan 2014 Word Chapter 6: Summary: What this thesis adds to the field and future work

particularly with regards to arrhythmia.

Finally, these analyses were performed with conventional statistical methodology and yielded

novel and informative findings. Going forwards, working with UK (University of Oxford)

and international collaborators (University of California, San Francsico), I am now exploring

using unsupervised machine learning on this dataset to detect clinically or biologically

important relationships within the data that may offer novel prognostic phenotypes(415).

Limitations

Whilst much has been learnt from this genotype and phenotype study in DCM, the study is

limited by its cross sectional nature, with phenotyping at only one timepoint in the course of a

complex disease. The information gleaned, whilst extensive, may still be insufficient to

identify risk for multi-level disease outcomes in DCM. In addition, whilst risk prediction has

its role, the field is in particular need of therapies which modulate risk by targeting the

upstream drivers of adverse outcomes. Much remains to be explored regarding factors

influencing disease onset and progression of genetic DCM, epigenetics and environmental

factors not evaluated in this thesis.

Conclusions

The work in this thesis refines our current understanding of the genetics of DCM and

demonstrates the utility of genetic, advanced imaging, and phenotypic analysis for

stratification of patients with DCM.

355 Template by Friedman & Morgan 2014 Word Chapter 7: References

7 REFERENCES

1. Elliott P, Andersson B, Arbustini E et al. Classification of the cardiomyopathies: a position statement from the European Society Of Cardiology Working Group on Myocardial and Pericardial Diseases. Eur Heart J 2008;29:270-276. 2. Pinto YM, Elliott PM, Arbustini E et al. Proposal for a revised definition of dilated cardiomyopathy, hypokinetic non-dilated cardiomyopathy, and its implications for clinical practice: a position statement of the ESC working group on myocardial and pericardial diseases. Eur Heart J 2016;37:1850-1858. 3. Richardson P, McKenna W, Bristow M et al. Report of the 1995 World Health Organization/International Society and Federation of Cardiology Task Force on the Definition and Classification of cardiomyopathies. Circulation 1996;93:841-842. 4. Codd MB, Sugrue DD, Gersh BJ, Melton LJ, 3rd. Epidemiology of idiopathic dilated and hypertrophic cardiomyopathy. A population-based study in Olmsted County, Minnesota, 1975-1984. Circulation 1989;80:564-572. 5. Hershberger RE, Hedges DJ, Morales A. Dilated cardiomyopathy: the complexity of a diverse genetic architecture. Nat Rev Cardiol 2013;10:531-547. 6. International Society of Heart and Lung Transplantation Quarterly Report. 2015 ed. 7. Maron BJ, Towbin JA, Thiene G et al. Contemporary definitions and classification of the cardiomyopathies: an American Heart Association Scientific Statement from the Council on Clinical Cardiology, Heart Failure and Transplantation Committee; Quality of Care and Outcomes Research and Functional Genomics and Translational Biology Interdisciplinary Working Groups; and Council on Epidemiology and Prevention. Circulation 2006;113:1807-1816. 8. Elliott P, Andersson B, Arbustini E et al. Classification of the cardiomyopathies: a position statement from the european society of cardiology working group on myocardial and pericardial diseases. European Heart Journal 2008;29:270-276. 9. Henry WL, Gardin JM, Ware JH. Echocardiographic measurements in normal subjects from infancy to old age. Circulation 1980;62:1054-1061. 10. Caforio AL, Pankuweit S, Arbustini E et al. Current state of knowledge on aetiology, diagnosis, management, and therapy of myocarditis: a position statement of the European Society of Cardiology Working Group on Myocardial and Pericardial Diseases. Eur Heart J 2013;34:2636-2648. 11. Cole GD, Dhutia NM, Shun-Shin MJ et al. Defining the real-world reproducibility of visual grading of left ventricular function and visual estimation of left ventricular ejection fraction: impact of image quality, experience and accreditation. Int J Cardiovasc Imaging 2015;31:1303-1314. 12. Kober L, Thune JJ, Nielsen JC et al. Defibrillator Implantation in Patients with Nonischemic Systolic Heart Failure. N Engl J Med 2016;375:1221-1230. 13. Gulati A, Jabbour A, Ismail TF et al. Association of fibrosis with mortality and sudden cardiac death in patients with nonischemic dilated cardiomyopathy. JAMA 2013;309:896-908. 14. Walsh R, Thomson KL, Ware JS et al. Reassessment of Mendelian gene pathogenicity using 7,855 cardiomyopathy cases and 60,706 reference samples. Genet Med 2017;19:192-203. 15. Lek M, Karczewski KJ, Minikel EV et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature 2016;536:285-291. 16. Petretta M, Pirozzi F, Sasso L, Paglia A, Bonaduce D. Review and metaanalysis of the frequency of familial dilated cardiomyopathy. Am J Cardiol 2011;108:1171-1176. 17. Mestroni L, Brun F, Spezzacatene A, Sinagra G, Taylor MR. Genetic Causes of Dilated Cardiomyopathy. Prog Pediatr Cardiol 2014;37:13-18. 18. Nouhravesh N, Ahlberg G, Ghouse J et al. Analyses of more than 60,000 exomes questions the role of numerous genes previously associated with dilated cardiomyopathy. Mol Genet Genomic Med 2016;4:617-623. 19. Haas J, Frese KS, Peil B et al. Atlas of the clinical genetics of human dilated cardiomyopathy, 2015. 20. Bezzina CR, Lahrouchi N, Priori SG. Genetics of Sudden Cardiac Death. Circulation Research 2015;116:1919-1936. 21. Giudicessi JR, Kullo IJ, Ackerman MJ. Precision Cardiovascular Medicine: State of Genetic Testing.

356 Template by Friedman & Morgan 2014 Word Chapter 7: References

Mayo Clinic Proceedings 2017;92:642-662. 22. Geisterfer-Lowrance AA, Kass S, Tanigawa G et al. A molecular basis for familial hypertrophic cardiomyopathy: a beta cardiac myosin heavy chain gene missense mutation. Cell 1990;62:999-1006. 23. Gibson G. Rare and common variants: twenty arguments. Nat Rev Genet 2012;13:135-145. 24. Whiffin N, Minikel E, Walsh R et al. Using high-resolution variant frequencies to empower clinical genome interpretation. Genet Med 2017;10.1038/gim.2017.26. 25. Frequency Filter: Using high-resolution variant frequencies to empower clinical genome interpretation. https://cardiodb.org/allelefrequencyapp/. Accessed:30.1.17. 26. Tayal U, Prasad S, Cook SA. Genetics and genomics of dilated cardiomyopathy and systolic heart failure. Genome Med 2017;9:20. 27. Sanger F, Nicklen S, Coulson AR. DNA sequencing with chain-terminating inhibitors. Proc Natl Acad Sci U S A 1977;74:5463-5467. 28. Lander ES, Linton LM, Birren B et al. Initial sequencing and analysis of the human genome. Nature 2001;409:860-921. 29. Begay RL, Tharp CA, Martin A et al. FLNC Gene Splice Mutations Cause Dilated Cardiomyopathy. JACC Basic Transl Sci 2016;1:344-359. 30. The Genomes Project C. A global reference for human genetic variation. Nature 2015;526:68-74. 31. Tennessen JA, Bigham AW, O'Connor TD et al. Evolution and functional impact of rare coding variation from deep sequencing of human exomes. Science 2012;337:64-69. 32. The UKKC. The UK10K project identifies rare variants in health and disease. Nature 2015;526:82-90. 33. Landrum MJ, Lee JM, Riley GR et al. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res 2014;42:D980-985. 34. Stenson PD, Mort M, Ball EV et al. The Human Gene Mutation Database: building a comprehensive mutation repository for clinical and molecular genetics, diagnostic testing and personalized genomic medicine. Hum Genet 2014;133:1-9. 35. Liu X, Wu C, Li C, Boerwinkle E. dbNSFP v3.0: A One-Stop Database of Functional Predictions and Annotations for Human Nonsynonymous and Splice-Site SNVs. Hum Mutat 2016;37:235-241. 36. Richards S, Aziz N, Bale S et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med 2015;17:405-423. 37. Ashley EA, Butte AJ, Wheeler MT et al. Clinical assessment incorporating a personal genome. Lancet 2010;375:1525-1535. 38. Amendola LM, Jarvik GP, Leo MC et al. Performance of ACMG-AMP Variant-Interpretation Guidelines among Nine Laboratories in the Clinical Sequencing Exploratory Research Consortium. Am J Hum Genet 2016;99:247. 39. Rehm HL, Berg JS, Brooks LD et al. ClinGen--the Clinical Genome Resource. N Engl J Med 2015;372:2235-2242. 40. Dalin MG, Engstrom PG, Ivarsson EG et al. Massive parallel sequencing questions the pathogenic role of missense variants in dilated cardiomyopathy. Int J Cardiol 2017;228:742-748. 41. Cahill TJ, Ashrafian H, Watkins H. Genetic cardiomyopathies causing heart failure. Circ Res 2013;113:660-675. 42. Robinson P, Griffiths PJ, Watkins H, Redwood CS. Dilated and Hypertrophic Cardiomyopathy Mutations in Troponin and α-Tropomyosin Have Opposing Effects on the Calcium Affinity of Cardiac Thin Filaments. Circulation Research 2007;101:1266-1273. 43. Olson TM, Michels VV, Thibodeau SN, Tai YS, Keating MT. Actin mutations in dilated cardiomyopathy, a heritable form of heart failure. Science 1998;280:750-752. 44. Daehmlow S, Erdmann J, Knueppel T et al. Novel mutations in sarcomeric protein genes in dilated cardiomyopathy. Biochem Biophys Res Commun 2002;298:116-120. 45. Hershberger RE, Pinto JR, Parks SB et al. Clinical and functional characterization of TNNT2 mutations identified in patients with dilated cardiomyopathy. Circ Cardiovasc Genet 2009;2:306-313. 46. Pinto JR, Siegfried JD, Parvatiyar MS et al. Functional characterization of TNNC1 rare variants identified in dilated cardiomyopathy. J Biol Chem 2011;286:34404-34412. 47. Fatkin D, MacRae C, Sasaki T et al. Missense mutations in the rod domain of the lamin A/C gene as causes of dilated cardiomyopathy and conduction-system disease. N Engl J Med 1999;341:1715-1724. 48. Arbustini E, Pilotto A, Repetto A et al. Autosomal dominant dilated cardiomyopathy with atrioventricular block: a lamin A/C defect-related disease. J Am Coll Cardiol 2002;39:981-990. 49. van Rijsingen IA, Arbustini E, Elliott PM et al. Risk factors for malignant ventricular arrhythmias in lamin a/c mutation carriers a European cohort study. J Am Coll Cardiol 2012;59:493-500. 50. Parent JJ, Towbin JA, Jefferies JL. Left ventricular noncompaction in a family with lamin a/c gene mutation. Tex Heart Inst J 2015;42:73-76.

357 Template by Friedman & Morgan 2014 Word Chapter 7: References

51. Norton N, Li D, Rieder MJ et al. Genome-wide studies of copy number variation and exome sequencing identify rare variants in BAG3 as a cause of dilated cardiomyopathy. Am J Hum Genet 2011;88:273-282. 52. Villard E, Perret C, Gary F et al. A genome-wide association study identifies two loci associated with heart failure due to dilated cardiomyopathy. Eur Heart J 2011;32:1065-1076. 53. Guo W, Schafer S, Greaser ML et al. RBM20, a gene for hereditary cardiomyopathy, regulates titin splicing. Nat Med 2012;18:766-773. 54. Beqqali A, Bollen IA, Rasmussen TB et al. A mutation in the glutamate-rich region of RNA-binding motif protein 20 causes dilated cardiomyopathy through missplicing of titin and impaired Frank- Starling mechanism. Cardiovasc Res 2016;112:452-463. 55. Brauch KM, Karst ML, Herron KJ et al. Mutations in ribonucleic acid binding protein gene cause familial dilated cardiomyopathy. J Am Coll Cardiol 2009;54:930-941. 56. Li D, Morales A, Gonzalez-Quintana J et al. Identification of novel mutations in RBM20 in patients with dilated cardiomyopathy. Clin Transl Sci 2010;3:90-97. 57. Refaat MM, Lubitz SA, Makino S et al. Genetic variation in the alternative splicing regulator RBM20 is associated with dilated cardiomyopathy. Heart Rhythm 2012;9:390-396. 58. Arndt AK, Schafer S, Drenckhahn JD et al. Fine mapping of the 1p36 deletion syndrome identifies mutation of PRDM16 as a cause of cardiomyopathy. Am J Hum Genet 2013;93:67-77. 59. de Leeuw N, Houge G. Loss of PRDM16 Is Unlikely to Cause Cardiomyopathy in 1p36 Deletion Syndrome. The American Journal of Human Genetics 2014;94:153-154. 60. Buyandelger B, Mansfield C, Kostin S et al. ZBTB17 (MIZ1) Is Important for the Cardiac Stress Response and a Novel Candidate Gene for Cardiomyopathy and Heart Failure. Circulation: Cardiovascular Genetics 2015;8:643-652. 61. Knoll R, Hoshijima M, Hoffman HM et al. The cardiac mechanical stretch sensor machinery involves a Z disc complex that is defective in a subset of human dilated cardiomyopathy. Cell 2002;111:943-955. 62. Schmitt JP, Kamisago M, Asahi M et al. Dilated cardiomyopathy and heart failure caused by a mutation in phospholamban. Science 2003;299:1410-1413. 63. Haghighi K, Kolokathis F, Gramolini AO et al. A mutation in the human phospholamban gene, deleting arginine 14, results in lethal, hereditary cardiomyopathy. Proc Natl Acad Sci U S A 2006;103:1388-1393. 64. DeWitt MM, MacLeod HM, Soliven B, McNally EM. Phospholamban R14 deletion results in late- onset, mild, hereditary dilated cardiomyopathy. J Am Coll Cardiol 2006;48:1396-1398. 65. van der Zwaag PA, van Rijsingen IA, Asimaki A et al. Phospholamban R14del mutation in patients diagnosed with dilated cardiomyopathy or arrhythmogenic right ventricular cardiomyopathy: evidence supporting the concept of arrhythmogenic cardiomyopathy. Eur J Heart Fail 2012;14:1199-1207. 66. van Spaendonck-Zwarts KY, van Rijsingen IA, van den Berg MP et al. Genetic analysis in 418 index patients with idiopathic dilated cardiomyopathy: overview of 10 years' experience. Eur J Heart Fail 2013;15:628-636. 67. Pugh TJ, Kelly MA, Gowrisankar S et al. The landscape of genetic variation in dilated cardiomyopathy as surveyed by clinical DNA sequencing. Genet Med 2014;16:601-608. 68. van der Zwaag PA, van Rijsingen IA, de Ruiter R et al. Recurrent and founder mutations in the Netherlands-Phospholamban p.Arg14del mutation causes arrhythmogenic cardiomyopathy. Neth Heart J 2013;21:286-293. 69. Zhang XL, Dai N, Tang K et al. GATA5 loss-of-function mutation in familial dilated cardiomyopathy. International Journal of Molecular Medicine 2015;35:763-770. 70. Zhao CM, Bing-Sun, Song HM et al. TBX20 loss-of-function mutation associated with familial dilated cardiomyopathy. Clinical Chemistry and Laboratory Medicine 2016;54:325-332. 71. Zhou W, Zhao L, Jiang JQ et al. A novel TBX5 loss-of-function mutation associated with sporadic dilated cardiomyopathy. International Journal of Molecular Medicine 2015;36:282-288. 72. Xu L, Zhao L, Yuan F et al. GATA6 loss-of-function mutations contribute to familial dilated cardiomyopathy. International Journal of Molecular Medicine 2014;34:1315-1322. 73. Zhao L, Xu JH, Xu WJ et al. A novel GATA4 loss-of-function mutation responsible for familial dilated cardiomyopathy. International Journal of Molecular Medicine 2014;33:654-660. 74. Yuan F, Qiu XB, Li RG et al. A novel NKX2-5 loss-of-function mutation predisposes to familial dilated cardiomyopathy and arrhythmias. International Journal of Molecular Medicine 2015;35:478- 486. 75. Akinrinade O, Ollila L, Vattulainen S et al. Genetics and genotype-phenotype correlations in Finnish patients with dilated cardiomyopathy. Eur Heart J 2015;36:2327-2337. 76. Garcia-Pavia P, Syrris P, Salas C et al. Desmosomal protein gene mutations in patients with idiopathic dilated cardiomyopathy undergoing cardiac transplantation: a clinicopathological study. Heart

358 Template by Friedman & Morgan 2014 Word Chapter 7: References

2011;97:1744-1752. 77. Christensen AH, Kamstrup PR, Gandjbakhch E et al. Plakophilin-2 c.419C > T and risk of heart failure and arrhythmias in the general population. European Journal of Human Genetics 2016;24:732-738. 78. Ortiz-Genga MF, Cuenca S, Dal Ferro M et al. Truncating FLNC Mutations Are Associated With High-Risk Dilated and Arrhythmogenic Cardiomyopathies. J Am Coll Cardiol 2016;68:2440-2451. 79. Herman DS, Lam L, Taylor MR et al. Truncations of titin causing dilated cardiomyopathy. N Engl J Med 2012;366:619-628. 80. LeWinter MM, Granzier H. Cardiac titin: a multifunctional giant. Circulation 2010;121:2137-2145. 81. LeWinter MM, Granzier HL. Cardiac Titin and Heart Disease. Journal of cardiovascular pharmacology 2014;63:207-212. 82. LeWinter MM, Zile MR. Could Modification of Titin Contribute to an Answer for Heart Failure With Preserved Ejection Fraction? Circulation 2016;134:1100-1104. 83. Gerull B, Gramlich M, Atherton J et al. Mutations of TTN, encoding the giant muscle filament titin, cause familial dilated cardiomyopathy. Nat Genet 2002;30:201-204. 84. Roberts AM, Ware JS, Herman DS et al. Integrated allelic, transcriptional, and phenomic dissection of the cardiac effects of titin truncations in health and disease. Sci Transl Med 2015;7:270ra276. 85. Jansweijer JA, Nieuwhof K, Russo F et al. Truncating titin mutations are associated with a mild and treatable form of dilated cardiomyopathy. Eur J Heart Fail 2017;19:512-521. 86. Franaszczyk M, Chmielewski P, Truszkowska G et al. Titin Truncating Variants in Dilated Cardiomyopathy - Prevalence and Genotype-Phenotype Correlations. PLoS One 2017;12:e0169007. 87. Akinrinade O, Koskenvuo JW, Alastalo TP. Prevalence of Titin Truncating Variants in General Population. PLoS One 2015;10:e0145284. 88. Schafer S, de Marvao A, Adami E et al. Titin-truncating variants affect heart function in disease cohorts and the general population. Nat Genet 2017;49:46-53. 89. Roncarati R, Viviani Anselmi C, Krawitz P et al. Doubly heterozygous LMNA and TTN mutations revealed by exome sequencing in a severe form of dilated cardiomyopathy. Eur J Hum Genet 2013;21:1105-1111. 90. Guo W, Pleitner JM, Saupe KW, Greaser ML. Pathophysiological defects and transcriptional profiling in the RBM20-/- rat model. PLoS One 2013;8:e84281. 91. Egger G, Liang G, Aparicio A, Jones PA. Epigenetics in human disease and prospects for epigenetic therapy. Nature 2004;429:457-463. 92. Ware JS, Li J, Mazaika E et al. Shared Genetic Predisposition in Peripartum and Dilated Cardiomyopathies. N Engl J Med 2016;374:233-241. 93. Linschoten M, Teske AJ, Baas AF et al. Truncating Titin (TTN) Variants in Chemotherapy-Induced Cardiomyopathy. J Card Fail 2017;23:476-479. 94. Gramlich M, Michely B, Krohne C et al. Stress-induced dilated cardiomyopathy in a knock-in mouse model mimicking human titin-based disease. J Mol Cell Cardiol 2009;47:352-358. 95. Hazebroek MR, Moors S, Dennert R et al. Prognostic Relevance of Gene-Environment Interactions in Patients With Dilated Cardiomyopathy: Applying the MOGE(S) Classification. J Am Coll Cardiol 2015;66:1313-1323. 96. Whitman IR, Agarwal V, Nah G et al. Alcohol Abuse and Cardiac Disease. J Am Coll Cardiol 2017;69:13-24. 97. Guzzo-Merello G, Segovia J, Dominguez F et al. Natural history and prognostic factors in alcoholic cardiomyopathy. JACC Heart Fail 2015;3:78-86. 98. Milic S, Lulic D, Stimac D, Ruzic A, Zaputovic L. Cardiac manifestations in alcoholic liver disease. Postgrad Med J 2016;92:235-239. 99. Bell S, Daskalopoulou M, Rapsomaniki E et al. Association between clinically recorded alcohol consumption and initial presentation of 12 cardiovascular diseases: population based cohort study using linked health records. BMJ 2017;356:j909. 100. Fernandez-Sola J, Nicolas JM, Oriola J et al. Angiotensin-converting enzyme gene polymorphism is associated with vulnerability to alcoholic cardiomyopathy. Ann Intern Med 2002;137:321-326. 101. Piano MR, Phillips SA. Alcoholic cardiomyopathy: Pathophysiologic insights. Cardiovascular toxicology 2014;14:291-308. 102. Shellock FG, Crues JV. MR procedures: biologic effects, safety, and patient care. Radiology 2004;232:635-652. 103. Pennell DJ. Cardiovascular Magnetic Resonance. Circulation 2010;121:692-705. 104. Fiechter M, Stehli J, Fuchs TA et al. Impact of cardiac magnetic resonance imaging on human lymphocyte DNA integrity. Eur Heart J 2013;34:2340-2345. 105. Lancellotti P, Nchimi A, Delierneux C et al. Biological Effects of Cardiac Magnetic Resonance on Human Blood Cells. Circ Cardiovasc Imaging 2015;8:e003697.

359 Template by Friedman & Morgan 2014 Word Chapter 7: References

106. Hill MA, O'Neill P, McKenna WG. Comments on potential health effects of MRI-induced DNA lesions: quality is more important to consider than quantity. Eur Heart J Cardiovasc Imaging 2016;17:1230-1238. 107. Assomull RG, Lyne JC, Keenan N et al. The role of cardiovascular magnetic resonance in patients presenting with chest pain, raised troponin, and unobstructed coronary arteries. Eur Heart J 2007;28:1242-1249. 108. McCrohon JA, Moon JC, Prasad SK et al. Differentiation of heart failure related to dilated cardiomyopathy and coronary artery disease using gadolinium-enhanced cardiovascular magnetic resonance. Circulation 2003;108:54-59. 109. Anderson LJ, Holden S, Davis B et al. Cardiovascular T2-star (T2*) magnetic resonance for the early diagnosis of myocardial iron overload. Eur Heart J 2001;22:2171-2179. 110. Grimm W, Alter P, Maisch B. Arrhythmia risk stratification with regard to prophylactic implantable defibrillator therapy in patients with dilated cardiomyopathy. Results of MACAS, DEFINITE, and SCD-HeFT. Herz 2004;29:348-352. 111. Authors/Task Force M, Priori SG, Blomstrom-Lundqvist C et al. 2015 ESC Guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: The Task Force for the Management of Patients with Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death of the European Society of Cardiology (ESC)Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC). Europace 2015;17:1601-1687. 112. Russo AM, Stainback RF, Bailey SR et al. ACCF/HRS/AHA/ASE/HFSA/SCAI/SCCT/SCMR 2013 appropriate use criteria for implantable cardioverter-defibrillators and cardiac resynchronization therapy: a report of the American College of Cardiology Foundation appropriate use criteria task force, Heart Rhythm Society, American Heart Association, American Society of Echocardiography, Heart Failure Society of America, Society for Cardiovascular Angiography and Interventions, Society of Cardiovascular Computed Tomography, and Society for Cardiovascular Magnetic Resonance. J Am Coll Cardiol 2013;61:1318-1368. 113. Gulati A, Ismail TF, Jabbour A et al. The prevalence and prognostic significance of right ventricular systolic dysfunction in nonischemic dilated cardiomyopathy. Circulation 2013;128:1623-1633. 114. Mikami Y, Jolly U, Heydari B et al. Right Ventricular Ejection Fraction Is Incremental to Left Ventricular Ejection Fraction for the Prediction of Future Arrhythmic Events in Patients With Systolic Dysfunction. Circ Arrhythm Electrophysiol 2017;10:e004067. 115. Buser PT, Auffermann W, Holt WW et al. Noninvasive evaluation of global left ventricular function with use of cine nuclear magnetic resonance. J Am Coll Cardiol 1989;13:1294-1300. 116. Maceira AM, Prasad SK, Khan M, Pennell DJ. Normalized left ventricular systolic and diastolic function by steady state free precession cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2006;8:417-426. 117. Globits S, Pacher R, Frank H et al. Comparative assessment of right ventricular volumes and ejection fraction by thermodilution and magnetic resonance imaging in dilated cardiomyopathy. Cardiology 1995;86:67-72. 118. Maceira AM, Prasad SK, Khan M, Pennell DJ. Reference right ventricular systolic and diastolic function normalized to age, gender and body surface area from steady-state free precession cardiovascular magnetic resonance. Eur Heart J 2006;27:2879-2888. 119. Grothues F, Moon JC, Bellenger NG et al. Interstudy reproducibility of right ventricular volumes, function, and mass with cardiovascular magnetic resonance. Am Heart J 2004;147:218-223. 120. Grothues F, Smith GC, Moon JC et al. Comparison of interstudy reproducibility of cardiovascular magnetic resonance with two-dimensional echocardiography in normal subjects and in patients with heart failure or left ventricular hypertrophy. Am J Cardiol 2002;90:29-34. 121. Karamitsos TD, Hudsmith LE, Selvanayagam JB, Neubauer S, Francis JM. Operator induced variability in left ventricular measurements with cardiovascular magnetic resonance is improved after training. J Cardiovasc Magn Reson 2007;9:777-783. 122. Lang RM, Bierig M, Devereux RB et al. Recommendations for chamber quantification: a report from the American Society of Echocardiography's Guidelines and Standards Committee and the Chamber Quantification Writing Group, developed in conjunction with the European Association of Echocardiography, a branch of the European Society of Cardiology. J Am Soc Echocardiogr 2005;18:1440-1463. 123. Anderson JL, Horne BD, Pennell DJ. Atrial dimensions in health and left ventricular disease using cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2005;7:671-675. 124. Whitlock M, Garg A, Gelow J, Jacobson T, Broberg C. Comparison of left and right atrial volume by echocardiography versus cardiac magnetic resonance imaging using the area-length method. Am J Cardiol 2010;106:1345-1350.

360 Template by Friedman & Morgan 2014 Word Chapter 7: References

125. Pritchett AM, Mahoney DW, Jacobsen SJ et al. Diastolic dysfunction and left atrial volume: a population-based study. J Am Coll Cardiol 2005;45:87-92. 126. Therkelsen SK, Groenning BA, Svendsen JH, Jensen GB. Atrial and ventricular volume and function in persistent and permanent atrial fibrillation, a magnetic resonance imaging study. J Cardiovasc Magn Reson 2005;7:465-473. 127. Gulati A, Ismail TF, Jabbour A et al. Clinical utility and prognostic value of left atrial volume assessment by cardiovascular magnetic resonance in non-ischaemic dilated cardiomyopathy. Eur J Heart Fail 2013;15:660-670. 128. Chan KM, Wage R, Symmonds K et al. Towards comprehensive assessment of mitral regurgitation using cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2008;10:61. 129. Kizilbash AM, Hundley WG, Willett DL et al. Comparison of quantitative Doppler with magnetic resonance imaging for assessment of the severity of mitral regurgitation. Am J Cardiol 1998;81:792- 795. 130. Kon MW, Myerson SG, Moat NE, Pennell DJ. Quantification of regurgitant fraction in mitral regurgitation by cardiovascular magnetic resonance: comparison of techniques. J Heart Valve Dis 2004;13:600-607. 131. Stolfo D, Merlo M, Pinamonti B et al. Early improvement of functional mitral regurgitation in patients with idiopathic dilated cardiomyopathy. Am J Cardiol 2015;115:1137-1143. 132. Bocchi EA, Bellotti G, Moreira LF et al. Prognostic indicators of one-year outcome after cardiomyoplasty for idiopathic dilated cardiomyopathy. Am J Cardiol 1994;73:604-608. 133. Doherty NE, 3rd, Seelos KC, Suzuki J et al. Application of cine nuclear magnetic resonance imaging for sequential evaluation of response to angiotensin-converting enzyme inhibitor therapy in dilated cardiomyopathy. J Am Coll Cardiol 1992;19:1294-1302. 134. Bellenger NG, Davies LC, Francis JM, Coats AJ, Pennell DJ. Reduction in sample size for studies of remodeling in heart failure by the use of cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2000;2:271-278. 135. Davies MJ, McKenna WJ. Dilated cardiomyopathy: an introduction to pathology and pathogenesis. Br Heart J 1994;72:S24. 136. Olivotto I, d'Amati G, Basso C et al. Defining phenotypes and disease progression in sarcomeric cardiomyopathies: contemporary role of clinical investigations. Cardiovasc Res 2015;105:409-423. 137. Raman SV, Basso C, Tandri H, Taylor MR. Imaging phenotype vs genotype in nonhypertrophic heritable cardiomyopathies: dilated cardiomyopathy and arrhythmogenic right ventricular cardiomyopathy. Circ Cardiovasc Imaging 2010;3:753-765. 138. Eitel I, Friedrich MG. T2-weighted cardiovascular magnetic resonance in acute cardiac disease. J Cardiovasc Magn Reson 2011;13:13. 139. Friedrich MG, Sechtem U, Schulz-Menger J et al. Cardiovascular magnetic resonance in myocarditis: A JACC White Paper. J Am Coll Cardiol 2009;53:1475-1487. 140. Anderson KR, Sutton MG, Lie JT. Histopathological types of cardiac fibrosis in myocardial disease. J Pathol 1979;128:79-85. 141. Assomull RG, Prasad SK, Lyne J et al. Cardiovascular magnetic resonance, fibrosis, and prognosis in dilated cardiomyopathy. J Am Coll Cardiol 2006;48:1977-1985. 142. Moon JC, Reed E, Sheppard MN et al. The histologic basis of late gadolinium enhancement cardiovascular magnetic resonance in hypertrophic cardiomyopathy. J Am Coll Cardiol 2004;43:2260- 2264. 143. Kim RJ, Fieno DS, Parrish TB et al. Relationship of MRI delayed contrast enhancement to irreversible injury, infarct age, and contractile function. Circulation 1999;100:1992-2002. 144. Schelbert EB, Hsu LY, Anderson SA et al. Late gadolinium-enhancement cardiac magnetic resonance identifies postinfarction myocardial fibrosis and the border zone at the near cellular level in ex vivo rat heart. Circ Cardiovasc Imaging 2010;3:743-752. 145. Iles LM, Ellims AH, Llewellyn H et al. Histological validation of cardiac magnetic resonance analysis of regional and diffuse interstitial myocardial fibrosis. Eur Heart J Cardiovasc Imaging 2015;16:14-22. 146. Halliday BP, Gulati A, Ali A et al. Association Between Midwall Late Gadolinium Enhancement and Sudden Cardiac Death in Patients With Dilated Cardiomyopathy and Mild and Moderate Left Ventricular Systolic Dysfunction. Circulation 2017;135:2106-2115. 147. Memon S, Ganga HV, Kluger J. Late Gadolinium Enhancement in Patients with Non-Ischemic Dilated Cardiomyopathy. Pacing Clin Electrophysiol 2016;39:731-747. 148. Neilan TG, Coelho-Filho OR, Danik SB et al. CMR quantification of myocardial scar provides additive prognostic information in nonischemic cardiomyopathy. JACC Cardiovasc Imaging 2013;6:944-954. 149. Brooks A, Schinde V, Bateman AC, Gallagher PJ. Interstitial fibrosis in the dilated non-ischaemic myocardium. Heart 2003;89:1255-1256.

361 Template by Friedman & Morgan 2014 Word Chapter 7: References

150. From AM, Maleszewski JJ, Rihal CS. Current Status of Endomyocardial Biopsy. Mayo Clinic Proceedings 2011;86:1095-1102. 151. Messroghli DR, Radjenovic A, Kozerke S et al. Modified Look-Locker inversion recovery (MOLLI) for high-resolution T1 mapping of the heart. Magn Reson Med 2004;52:141-146. 152. Piechnik SK, Ferreira VM, Dall'Armellina E et al. Shortened Modified Look-Locker Inversion recovery (ShMOLLI) for clinical myocardial T1-mapping at 1.5 and 3 T within a 9 heartbeat breathhold. J Cardiovasc Magn Reson 2010;12:69. 153. Haaf P, Garg P, Messroghli DR et al. Cardiac T1 Mapping and Extracellular Volume (ECV) in clinical practice: a comprehensive review. J Cardiovasc Magn Reson 2016;18:89. 154. Sibley CT, Noureldin RA, Gai N et al. T1 Mapping in cardiomyopathy at cardiac MR: comparison with endomyocardial biopsy. Radiology 2012;265:724-732. 155. Wong TC, Piehler K, Meier CG et al. Association between extracellular matrix expansion quantified by cardiovascular magnetic resonance and short-term mortality. Circulation 2012;126:1206-1216. 156. aus dem Siepen F, Buss SJ, Messroghli D et al. T1 mapping in dilated cardiomyopathy with cardiac magnetic resonance: quantification of diffuse myocardial fibrosis and comparison with endomyocardial biopsy. Eur Heart J Cardiovasc Imaging 2015;16:210-216. 157. Karamitsos TD, Piechnik SK, Banypersad SM et al. Noncontrast T1 mapping for the diagnosis of cardiac amyloidosis. JACC Cardiovasc Imaging 2013;6:488-497. 158. Bull S, White SK, Piechnik SK et al. Human non-contrast T1 values and correlation with histology in diffuse fibrosis. Heart 2013;99:932-937. 159. Puntmann VO, Voigt T, Chen Z et al. Native T1 mapping in differentiation of normal myocardium from diffuse disease in hypertrophic and dilated cardiomyopathy. JACC Cardiovasc Imaging 2013;6:475-484. 160. Puntmann VO, Carr-White G, Jabbour A et al. T1-Mapping and Outcome in Nonischemic Cardiomyopathy: All-Cause Mortality and Heart Failure. JACC Cardiovasc Imaging 2016;9:40-50. 161. Dass S, Holloway CJ, Cochlin LE et al. No Evidence of Myocardial Oxygen Deprivation in Nonischemic Heart Failure. Circ Heart Fail 2015;8:1088-1093. 162. Neglia D, Michelassi C, Trivieri MG et al. Prognostic role of myocardial blood flow impairment in idiopathic left ventricular dysfunction. Circulation 2002;105:186-193. 163. Rigo F, Gherardi S, Galderisi M et al. The prognostic impact of coronary flow-reserve assessed by Doppler echocardiography in non-ischaemic dilated cardiomyopathy. Eur Heart J 2006;27:1319-1323. 164. Buss SJ, Breuninger K, Lehrke S et al. Assessment of myocardial deformation with cardiac magnetic resonance strain imaging improves risk stratification in patients with dilated cardiomyopathy. Eur Heart J Cardiovasc Imaging 2015;16:307-315. 165. Lakdawala NK, Thune JJ, Colan SD et al. Subtle abnormalities in contractile function are an early manifestation of sarcomere mutations in dilated cardiomyopathy. Circ Cardiovasc Genet 2012;5:503- 510. 166. Kawel-Boehm N, Maceira A, Valsangiacomo-Buechel E et al. Normal values for cardiovascular magnetic resonance in adults and children. Journal of Cardiovascular Magnetic Resonance 2015;17:29. 167. Reichek N. MRI myocardial tagging. J Magn Reson Imaging 1999;10:609-616. 168. Wehner GJ, Suever JD, Haggerty CM et al. Validation of in vivo 2D displacements from spiral cine DENSE at 3T. J Cardiovasc Magn Reson 2015;17:5. 169. Auger DA, Zhong X, Epstein FH, Spottiswoode BS. Mapping right ventricular myocardial mechanics using 3D cine DENSE cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2012;14:4. 170. Gilliam AD, Epstein FH. Automated motion estimation for 2-D cine DENSE MRI. IEEE Trans Med Imaging 2012;31:1669-1681. 171. Matsumoto K, Tanaka H, Kaneko A et al. Contractile reserve assessed by three-dimensional global circumferential strain as a predictor of cardiovascular events in patients with idiopathic dilated cardiomyopathy. J Am Soc Echocardiogr 2012;25:1299-1308. 172. Ramahi TM, Longo MD, Cadariu AR et al. Dobutamine-induced augmentation of left ventricular ejection fraction predicts survival of heart failure patients with severe non-ischaemic cardiomyopathy. Eur Heart J 2001;22:849-856. 173. Kitaoka H, Takata J, Yabe T et al. Low dose dobutamine stress echocardiography predicts the improvement of left ventricular systolic function in dilated cardiomyopathy. Heart 1999;81:523-527. 174. Naqvi TZ, Goel RK, Forrester JS, Siegel RJ. Myocardial contractile reserve on dobutamine echocardiography predicts late spontaneous improvement in cardiac function in patients with recent onset idiopathic dilated cardiomyopathy. J Am Coll Cardiol 1999;34:1537-1544. 175. Kasama S, Toyama T, Kumakura H et al. Myocardial contractile reserve determined by dobutamine stress Tc-99m tetrofosmin quantitative gated SPECT predicts late spontaneous improvement in cardiac function in patients with recent-onset dilated cardiomyopathy. J Nucl Cardiol 2003;10:607-614.

362 Template by Friedman & Morgan 2014 Word Chapter 7: References

176. Matsumura Y, Takata J, Kitaoka H et al. Low-dose dobutamine stress echocardiography predicts the improvement of left ventricular systolic function and long-term prognosis in patients with idiopathic dilated cardiomyopathy. J Med Ultrason (2001) 2006;33:17-22. 177. Nagaoka H, Isobe N, Kubota S et al. Myocardial contractile reserve as prognostic determinant in patients with idiopathic dilated cardiomyopathy without overt heart failure. Chest 1997;111:344-350. 178. Scrutinio D, Napoli V, Passantino A et al. Low-dose dobutamine responsiveness in idiopathic dilated cardiomyopathy: relation to exercise capacity and clinical outcome. Eur Heart J 2000;21:927-934. 179. Kasama S, Toyama T, Hoshizaki H et al. Dobutamine gated blood pool scintigraphy predicts the improvement of cardiac sympathetic nerve activity, cardiac function, and symptoms after treatment in patients with dilated cardiomyopathy. Chest 2002;122:542-548. 180. Matsumoto K, Tanaka H, Onishi A et al. Bi-ventricular contractile reserve offers an incremental prognostic value for patients with dilated cardiomyopathy. Eur Heart J Cardiovasc Imaging 2015;16:1213-1223. 181. Fowler MB, Laser JA, Hopkins GL, Minobe W, Bristow MR. Assessment of the beta-adrenergic receptor pathway in the intact failing human heart: progressive receptor down-regulation and subsensitivity to agonist response. Circulation 1986;74:1290-1302. 182. Raphael CE, Vassiliou V, Alpendurada F et al. Clinical value of cardiovascular magnetic resonance in patients with MR-conditional pacemakers. Eur Heart J Cardiovasc Imaging 2016;17:1178-1185. 183. Klein-Wiele O, Garmer M, Busch M et al. Cardiovascular magnetic resonance in patients with magnetic resonance conditional pacemaker systems at 1.5 T: influence of pacemaker related artifacts on image quality including first pass perfusion, aortic and mitral valve assessment, flow measurement, short tau inversion recovery and T1-weighted imaging. Int J Cardiovasc Imaging 2017;33:383-394. 184. Givertz MM, Mann DL. Epidemiology and natural history of recovery of left ventricular function in recent onset dilated cardiomyopathies. Curr Heart Fail Rep 2013;10:321-330. 185. Dec GW, Jr., Palacios IF, Fallon JT et al. Active myocarditis in the spectrum of acute dilated cardiomyopathies. Clinical features, histologic correlates, and clinical outcome. N Engl J Med 1985;312:885-890. 186. Cintron G, Johnson G, Francis G, Cobb F, Cohn JN. Prognostic significance of serial changes in left ventricular ejection fraction in patients with congestive heart failure. The V-HeFT VA Cooperative Studies Group. Circulation 1993;87:VI17-23. 187. Steimle AE, Stevenson LW, Fonarow GC, Hamilton MA, Moriguchi JD. Prediction of improvement in recent onset cardiomyopathy after referral for heart transplantation. J Am Coll Cardiol 1994;23:553- 559. 188. Matsumura Y, Hoshikawa-Nagai E, Kubo T et al. Left ventricular reverse remodeling in long-term (>12 years) survivors with idiopathic dilated cardiomyopathy. Am J Cardiol 2013;111:106-110. 189. Kubanek M, Sramko M, Maluskova J et al. Novel Predictors of Left Ventricular Reverse Remodeling in Individuals With Recent-Onset Dilated Cardiomyopathy. Journal of the American College of Cardiology 2013;61:54-63. 190. Merlo M, Stolfo D, Anzini M et al. Persistent Recovery of Normal Left Ventricular Function and Dimension in Idiopathic Dilated Cardiomyopathy During Long-Term Follow-up: Does Real Healing Exist? Journal of the American Heart Association 2015;4:e001504-e001504. 191. Dal Ferro M, Stolfo D, Altinier A et al. Association between mutation status and left ventricular reverse remodelling in dilated cardiomyopathy. Heart 2017;10.1136/heartjnl-2016-311017. 192. McNamara DM, Starling RC, Cooper LT et al. Clinical and demographic predictors of outcomes in recent onset dilated cardiomyopathy: results of the IMAC (Intervention in Myocarditis and Acute Cardiomyopathy)-2 study. J Am Coll Cardiol 2011;58:1112-1118. 193. Merlo M, Pyxaras SA, Pinamonti B et al. Prevalence and prognostic significance of left ventricular reverse remodeling in dilated cardiomyopathy receiving tailored medical treatment. J Am Coll Cardiol 2011;57:1468-1476. 194. Amorim S, Campelo M, Martins E et al. Prevalence, predictors and prognosis of ventricular reverse remodeling in idiopathic dilated cardiomyopathy. Rev Port Cardiol 2016;35:253-260. 195. Broch K, Murbraech K, Andreassen AK et al. Contemporary Outcome in Patients With Idiopathic Dilated Cardiomyopathy. Am J Cardiol 2015;116:952-959. 196. Ikeda Y, Inomata T, Fujita T et al. Cardiac fibrosis detected by magnetic resonance imaging on predicting time course diversity of left ventricular reverse remodeling in patients with idiopathic dilated cardiomyopathy. Heart Vessels 2016;31:1817-1825. 197. Moon J, Shim CY, Kim Y-J et al. Left Atrial Volume as a Predictor of Left Ventricular Functional Recovery in Patients With Dilated Cardiomyopathy and Absence of Delayed Enhancement in Cardiac Magnetic Resonance. Journal of Cardiac Failure 2016;22:265-271. 198. Okada M, Tanaka H, Matsumoto K et al. Subclinical myocardial dysfunction in patients with reverse-

363 Template by Friedman & Morgan 2014 Word Chapter 7: References

remodeled dilated cardiomyopathy. J Am Soc Echocardiogr 2012;25:726-732. 199. Wijesurendra RS, Liu A, Eichhorn C et al. Lone Atrial Fibrillation Is Associated With Impaired Left Ventricular Energetics That Persists Despite Successful Catheter Ablation. Circulation 2016;134:1068- 1081. 200. Mann DL, Barger PM, Burkhoff D. Myocardial recovery and the failing heart: myth, magic, or molecular target? J Am Coll Cardiol 2012;60:2465-2472. 201. Alahakoon PT, Binkley PF. Regression of normalized left ventricular systolic function after discontinuation of treatment in patients with dilated cardiomyopathy. Congest Heart Fail 2009;15:245- 246. 202. Park JS, Kim JW, Seo KW et al. Recurrence of left ventricular dysfunction in patients with restored idiopathic dilated cardiomyopathy. Clin Cardiol 2014;37:222-226. 203. Luk K, Bakhsh A, Giannetti N et al. Recovery in Patients With Dilated Cardiomyopathy With Loss-of- Function Mutations in the Titin Gene. JAMA Cardiol 2017;2:700-702. 204. Felkin LE, Walsh R, Ware JS et al. Recovery of Cardiac Function in Cardiomyopathy Caused by Titin Truncation. JAMA Cardiol 2016;1:234-235. 205. Gopinathannair R, Sullivan R, Olshansky B. Tachycardia-mediated cardiomyopathy: recognition and management. Curr Heart Fail Rep 2009;6:257-264. 206. Nerheim P, Birger-Botkin S, Piracha L, Olshansky B. Heart failure and sudden death in patients with tachycardia-induced cardiomyopathy and recurrent tachycardia. Circulation 2004;110:247-252. 207. Mason JW, O'Connell JB, Herskowitz A et al. A clinical trial of immunosuppressive therapy for myocarditis. The Myocarditis Treatment Trial Investigators. N Engl J Med 1995;333:269-275. 208. McNamara DM, Elkayam U, Alharethi R et al. Clinical Outcomes for Peripartum Cardiomyopathy in North America: Results of the IPAC Study (Investigations of Pregnancy-Associated Cardiomyopathy). J Am Coll Cardiol 2015;66:905-914. 209. Ewer MS, Vooletich MT, Durand JB et al. Reversibility of trastuzumab-related cardiotoxicity: new insights based on clinical course and response to medical treatment. J Clin Oncol 2005;23:7820-7826. 210. Breathett K, Allen LA, Udelson J, Davis G, Bristow M. Changes in Left Ventricular Ejection Fraction Predict Survival and Hospitalization in Heart Failure With Reduced Ejection Fraction. Circ Heart Fail 2016;9. 211. Kramer DG, Trikalinos TA, Kent DM et al. Quantitative evaluation of drug or device effects on ventricular remodeling as predictors of therapeutic effects on mortality in patients with heart failure and reduced ejection fraction: a meta-analytic approach. J Am Coll Cardiol 2010;56:392-406. 212. Brenyo A, Barsheshet A, Kutyifa V et al. Predictors of spontaneous reverse remodeling in mild heart failure patients with left ventricular dysfunction. Circ Heart Fail 2014;7:565-572. 213. Florea VG, Rector TS, Anand IS, Cohn JN. Heart Failure With Improved Ejection Fraction: Clinical Characteristics, Correlates of Recovery, and Survival: Results From the Valsartan Heart Failure Trial. Circ Heart Fail 2016;9. 214. Leong DP, Chakrabarty A, Shipp N et al. Effects of myocardial fibrosis and ventricular dyssynchrony on response to therapy in new-presentation idiopathic dilated cardiomyopathy: insights from cardiovascular magnetic resonance and echocardiography. Eur Heart J 2012;33:640-648. 215. Hoit BD. The Pythian Left Atrium: Predicting LV Reverse Remodeling and Cardiovascular Events in New-Onset Dilated Cardiomyopathy. Journal of Cardiac Failure 2016;22:280-282. 216. Blauwet LA, Delgado-Montero A, Ryo K et al. Right Ventricular Function in Peripartum Cardiomyopathy at Presentation Is Associated With Subsequent Left Ventricular Recovery and Clinical Outcomes. Circ Heart Fail 2016;9. 217. Zhang Y, Guallar E, Blasco-Colmenares E et al. Changes in Follow-Up Left Ventricular Ejection Fraction Associated With Outcomes in Primary Prevention Implantable Cardioverter-Defibrillator and Cardiac Resynchronization Therapy Device Recipients. J Am Coll Cardiol 2015;66:524-531. 218. Fuster V, Gersh BJ, Giuliani ER et al. The natural history of idiopathic dilated cardiomyopathy. The American Journal of Cardiology 1981;47:525-531. 219. Diaz RA, Obasohan A, Oakley CM. Prediction of outcome in dilated cardiomyopathy. British Heart Journal 1987;58:393-399. 220. Sugrue DD, Rodeheffer RJ, Codd MB et al. The clinical course of idiopathic dilated cardiomyopathy. A population-based study. Ann Intern Med 1992;117:117-123. 221. Komajda M, Jais JP, Reeves F et al. Factors predicting mortality in idiopathic dilated cardiomyopathy. Eur Heart J 1990;11:824-831. 222. Di Lenarda A, Secoli G, Perkan A et al. Changing mortality in dilated cardiomyopathy. The Heart Muscle Disease Study Group. Br Heart J 1994;72:S46-51. 223. Felker GM, Thompson RE, Hare JM et al. Underlying causes and long-term survival in patients with initially unexplained cardiomyopathy. N Engl J Med 2000;342:1077-1084.

364 Template by Friedman & Morgan 2014 Word Chapter 7: References

224. Dec GW, Fuster V. Idiopathic Dilated Cardiomyopathy. New England Journal of Medicine 1994;331:1564-1575. 225. Weintraub RG, Semsarian C, Macdonald P. Dilated cardiomyopathy. Lancet 2017;390:400-414. 226. Kadish A, Dyer A, Daubert JP et al. Prophylactic defibrillator implantation in patients with nonischemic dilated cardiomyopathy. N Engl J Med 2004;350:2151-2158. 227. Strickberger SA, Hummel JD, Bartlett TG et al. Amiodarone versus implantable cardioverter- defibrillator:randomized trial in patients with nonischemic dilated cardiomyopathy and asymptomatic nonsustained ventricular tachycardia--AMIOVIRT. J Am Coll Cardiol 2003;41:1707-1712. 228. Bansch D, Antz M, Boczor S et al. Primary prevention of sudden cardiac death in idiopathic dilated cardiomyopathy: the Cardiomyopathy Trial (CAT). Circulation 2002;105:1453-1458. 229. Bardy GH, Lee KL, Mark DB et al. Amiodarone or an implantable cardioverter-defibrillator for congestive heart failure. N Engl J Med 2005;352:225-237. 230. Theuns DAMJ, Smith T, Hunink MGM, Bardy GH, Jordaens L. Effectiveness of prophylactic implantation of cardioverter-defibrillators without cardiac resynchronization therapy in patients with ischaemic or non-ischaemic heart disease: a systematic review and meta-analysis. Europace 2010;12:1564-1570. 231. Packer DL, Prutkin JM, Hellkamp AS et al. Impact of implantable cardioverter-defibrillator, amiodarone, and placebo on the mode of death in stable patients with heart failure: analysis from the sudden cardiac death in heart failure trial. Circulation 2009;120:2170-2176. 232. Alexander PM, Daubeney PE, Nugent AW et al. Long-term outcomes of dilated cardiomyopathy diagnosed during childhood: results from a national population-based study of childhood cardiomyopathy. Circulation 2013;128:2039-2046. 233. Keogh AM, Baron DW, Hickie JB. Prognostic guides in patients with idiopathic or ischemic dilated cardiomyopathy assessed for cardiac transplantation. Am J Cardiol 1990;65:903-908. 234. Middlekauff HR, Stevenson WG, Stevenson LW, Saxon LA. Syncope in advanced heart failure: high risk of sudden death regardless of origin of syncope. J Am Coll Cardiol 1993;21:110-116. 235. Aleksova A, Sabbadini G, Merlo M et al. Natural history of dilated cardiomyopathy: from asymptomatic left ventricular dysfunction to heart failure--a subgroup analysis from the Trieste Cardiomyopathy Registry. J Cardiovasc Med (Hagerstown) 2009;10:699-705. 236. Ikram H, Williamson HG, Won M, Crozier IG, Wells EJ. The course of idiopathic dilated cardiomyopathy in New Zealand. Br Heart J 1987;57:521-527. 237. Kelly TL, Cremo R, Nielsen C, Shabetai R. Prediction of outcome in late-stage cardiomyopathy. Am Heart J 1990;119:1111-1121. 238. Franciosa JA, Wilen M, Ziesche S, Cohn JN. Survival in men with severe chronic left ventricular failure due to either coronary heart disease or idiopathic dilated cardiomyopathy. The American Journal of Cardiology 1983;51:831-836. 239. Pappachan JM, Varughese GI, Sriraman R, Arunagirinathan G. Diabetic cardiomyopathy: Pathophysiology, diagnostic evaluation and management. World J Diabetes 2013;4:177-189. 240. Buxton AE, Lee KL, Hafley GE et al. Limitations of ejection fraction for prediction of sudden death risk in patients with coronary artery disease: lessons from the MUSTT study. J Am Coll Cardiol 2007;50:1150-1157. 241. Meinertz T, Hofmann T, Kasper W et al. Significance of ventricular arrhythmias in idiopathic dilated cardiomyopathy. Am J Cardiol 1984;53:902-907. 242. Olshausen KV, Stienen U, Schwarz F, Kubler W, Meyer J. Long-term prognostic significance of ventricular arrhythmias in idiopathic dilated cardiomyopathy. Am J Cardiol 1988;61:146-151. 243. Gigli M, Stolfo D, Merlo M et al. Insights into mildly dilated cardiomyopathy: temporal evolution and long-term prognosis. Eur J Heart Fail 2017;19:531-539. 244. De Maria R, Gavazzi A, Caroli A et al. Ventricular arrhythmias in dilated cardiomyopathy as an independent prognostic hallmark. The American Journal of Cardiology 1992;69:1451-1457. 245. Gradman A, Deedwania P, Cody R et al. Predictors of total mortality and sudden death in mild to moderate heart failure. Captopril-Digoxin Study Group. J Am Coll Cardiol 1989;14:564-570. 246. Spezzacatene A, Sinagra G, Merlo M et al. Arrhythmogenic Phenotype in Dilated Cardiomyopathy: Natural History and Predictors of Life-Threatening Arrhythmias. J Am Heart Assoc 2015;4:e002149. 247. Merchant FM, Ikeda T, Pedretti RF et al. Clinical utility of microvolt T-wave alternans testing in identifying patients at high or low risk of sudden cardiac death. Heart Rhythm 2012;9:1256-1264 e1252. 248. Merchant FM, Zheng H, Bigger T et al. A combined anatomic and electrophysiologic substrate based approach for sudden cardiac death risk stratification. Am Heart J 2013;166:744-752. 249. Hohnloser SH, Klingenheben T, Bloomfield D, Dabbous O, Cohen RJ. Usefulness of microvolt T- wave alternans for prediction of ventricular tachyarrhythmic events in patients with dilated

365 Template by Friedman & Morgan 2014 Word Chapter 7: References

cardiomyopathy: results from a prospective observational study. J Am Coll Cardiol 2003;41:2220- 2224. 250. Goldberger JJ, Subačius H, Patel T, Cunnane R, Kadish AH. Sudden Cardiac Death Risk Stratification in Patients With Nonischemic Dilated Cardiomyopathy. Journal of the American College of Cardiology 2014;63:1879-1889. 251. Ponikowski P, Voors AA, Anker SD et al. 2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: The Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC)Developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J 2016;37:2129-2200. 252. Epstein AE, Dimarco JP, Ellenbogen KA et al. ACC/AHA/HRS 2008 Guidelines for device-based therapy of cardiac rhythm abnormalities. Heart Rhythm 2008;5:e1-62. 253. NICE. Implantable cardioverter defibrillators and cardiac resynchronisation therapy for arrhythmias and heart failure. NICE guideline TA314. 2014. 254. Multicenter Postinfarction Research G. Risk stratification and survival after myocardial infarction. N Engl J Med 1983;309:331-336. 255. Moss AJ, Zareba W, Hall WJ et al. Prophylactic implantation of a defibrillator in patients with myocardial infarction and reduced ejection fraction. N Engl J Med 2002;346:877-883. 256. Likoff MJ, Chandler SL, Kay HR. Clinical determinants of mortality in chronic congestive heart failure secondary to idiopathic dilated or to ischemic cardiomyopathy. Am J Cardiol 1987;59:634-638. 257. Lewis JF, Webber JD, Sutton LL, Chesoni S, Curry CL. Discordance in degree of right and left ventricular dilation in patients with dilated cardiomyopathy: Recognition and clinical implications. Journal of the American College of Cardiology 1993;21:649-654. 258. Mikami Y, Jolly U, Heydari B et al. Right Ventricular Ejection Fraction Is Incremental to Left Ventricular Ejection Fraction for the Prediction of Future Arrhythmic Events in Patients With Systolic Dysfunction. Circ Arrhythm Electrophysiol 2017;10. 259. Gavazzi A, De Maria R, Renosto G et al. The spectrum of left ventricular size in dilated cardiomyopathy: clinical correlates and prognostic implications. SPIC (Italian Multicenter Cardiomyopathy Study) Group. Am Heart J 1993;125:410-422. 260. Nazarian S, Bluemke DA, Lardo AC et al. Magnetic resonance assessment of the substrate for inducible ventricular tachycardia in nonischemic cardiomyopathy. Circulation 2005;112:2821-2825. 261. Lehrke S, Lossnitzer D, Schob M et al. Use of cardiovascular magnetic resonance for risk stratification in chronic heart failure: prognostic value of late gadolinium enhancement in patients with non- ischaemic dilated cardiomyopathy. Heart 2011;97:727-732. 262. Perazzolo Marra M, De Lazzari M, Zorzi A et al. Impact of the presence and amount of myocardial fibrosis by cardiac magnetic resonance on arrhythmic outcome and sudden cardiac death in nonischemic dilated cardiomyopathy. Heart Rhythm 2014;11:856-863. 263. Scott PA, Rosengarten JA, Curzen NP, Morgan JM. Late gadolinium enhancement cardiac magnetic resonance imaging for the prediction of ventricular tachyarrhythmic events: a meta-analysis. Eur J Heart Fail 2013;15:1019-1027. 264. Di Marco A, Anguera I, Schmitt M et al. Late Gadolinium Enhancement and the Risk for Ventricular Arrhythmias or Sudden Death in Dilated Cardiomyopathy: Systematic Review and Meta-Analysis. JACC Heart Fail 2017;5:28-38. 265. Schulz-Menger J, Bluemke DA, Bremerich J et al. Standardized image interpretation and post processing in cardiovascular magnetic resonance: Society for Cardiovascular Magnetic Resonance (SCMR) board of trustees task force on standardized post processing. J Cardiovasc Magn Reson 2013;15:35. 266. Chen Z, Sohal M, Voigt T et al. Myocardial tissue characterization by cardiac magnetic resonance imaging using T1 mapping predicts ventricular arrhythmia in ischemic and non-ischemic cardiomyopathy patients with implantable cardioverter-defibrillators. Heart Rhythm 2015;12:792-801. 267. Martinez-Rumayor A, Richards AM, Burnett JC, Januzzi JL, Jr. Biology of the natriuretic peptides. Am J Cardiol 2008;101:3-8. 268. Felker GM, Hasselblad V, Hernandez AF, O'Connor CM. Biomarker-guided therapy in chronic heart failure: a meta-analysis of randomized controlled trials. Am Heart J 2009;158:422-430. 269. van Kimmenade RR, Pinto YM, Bayes-Genis A et al. Usefulness of intermediate amino-terminal pro- brain natriuretic peptide concentrations for diagnosis and prognosis of acute heart failure. Am J Cardiol 2006;98:386-390. 270. Doust JA, Pietrzak E, Dobson A, Glasziou P. How well does B-type natriuretic peptide predict death and cardiac events in patients with heart failure: systematic review. BMJ 2005;330:625. 271. Berger R, Huelsman M, Strecker K et al. B-Type Natriuretic Peptide Predicts Sudden Death in Patients With Chronic Heart Failure. Circulation 2002;105:2392-2397.

366 Template by Friedman & Morgan 2014 Word Chapter 7: References

272. Fonarow GC, Peacock WF, Phillips CO et al. Admission B-type natriuretic peptide levels and in- hospital mortality in acute decompensated heart failure. J Am Coll Cardiol 2007;49:1943-1950. 273. Masson S, Latini R, Anand IS et al. Direct comparison of B-type natriuretic peptide (BNP) and amino- terminal proBNP in a large population of patients with chronic and symptomatic heart failure: the Valsartan Heart Failure (Val-HeFT) data. Clin Chem 2006;52:1528-1538. 274. Januzzi JL, Jr., Rehman SU, Mohammed AA et al. Use of amino-terminal pro-B-type natriuretic peptide to guide outpatient therapy of patients with chronic left ventricular systolic dysfunction. J Am Coll Cardiol 2011;58:1881-1889. 275. Nishiyama K, Tsutamoto T, Kawahara C et al. Relationship between biological variation in B-type natriuretic peptide and plasma renin concentration in stable outpatients with dilated cardiomyopathy. Circ J 2011;75:1897-1904. 276. Feng J, Schaus BJ, Fallavollita JA, Lee TC, Canty JM, Jr. Preload induces troponin I degradation independently of myocardial ischemia. Circulation 2001;103:2035-2037. 277. Miller WL, Hartman KA, Burritt MF, Burnett JC, Jr., Jaffe AS. Troponin, B-type natriuretic peptides and outcomes in severe heart failure: differences between ischemic and dilated cardiomyopathies. Clin Cardiol 2007;30:245-250. 278. Peacock WFt, De Marco T, Fonarow GC et al. Cardiac troponin and outcome in acute heart failure. N Engl J Med 2008;358:2117-2126. 279. Li X, Luo R, Jiang R et al. The prognostic use of serum concentrations of cardiac troponin-I, CK-MB and myoglobin in patients with idiopathic dilated cardiomyopathy. Heart Lung 2014;43:219-224. 280. Miettinen KH, Eriksson S, Magga J et al. Clinical significance of troponin I efflux and troponin autoantibodies in patients with dilated cardiomyopathy. J Card Fail 2008;14:481-488. 281. Rusconi P, Wilkinson JD, Sleeper LA et al. Differences in Presentation and Outcomes Between Children With Familial Dilated Cardiomyopathy and Children With Idiopathic Dilated Cardiomyopathy: A Report From the Pediatric Cardiomyopathy Registry Study Group. Circ Heart Fail 2017;10. 282. Disertori M, Quintarelli S, Mazzola S et al. The need to modify patient selection to improve the benefits of implantable cardioverter-defibrillator for primary prevention of sudden death in non- ischaemic dilated cardiomyopathy. Europace 2013;15:1693-1701. 283. Kayvanpour E, Sedaghat-Hamedani F, Amr A et al. Genotype-phenotype associations in dilated cardiomyopathy: meta-analysis on more than 8000 individuals. Clin Res Cardiol 2016;106:127-139. 284. Pasotti M, Klersy C, Pilotto A et al. Long-term outcome and risk stratification in dilated cardiolaminopathies. J Am Coll Cardiol 2008;52:1250-1260. 285. Posch MG, Perrot A, Geier C et al. Genetic deletion of arginine 14 in phospholamban causes dilated cardiomyopathy with attenuated electrocardiographic R amplitudes. Heart Rhythm 2009;6:480-486. 286. Merlo M, Sinagra G, Carniel E et al. Poor prognosis of rare sarcomeric gene variants in patients with dilated cardiomyopathy. Clin Transl Sci 2013;6:424-428. 287. Elliott P, O'Mahony C, Syrris P et al. Prevalence of Desmosomal Protein Gene Mutations in Patients With Dilated Cardiomyopathy. Circulation: Cardiovascular Genetics 2010;3:314-322. 288. McNair WP, Sinagra G, Taylor MR et al. SCN5A mutations associate with arrhythmic dilated cardiomyopathy and commonly localize to the voltage-sensing mechanism. J Am Coll Cardiol 2011;57:2160-2168. 289. Priori SG, Blomström-Lundqvist C, Mazzanti A et al. 2015 ESC Guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death. The Task Force for the Management of Patients with Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death of the European Society of Cardiology (ESC)Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC) 2015;36:2793-2867. 290. Golwala H, Bajaj NS, Arora G, Arora P. Implantable Cardioverter-Defibrillator for Nonischemic Cardiomyopathy: An Updated Meta-Analysis. Circulation 2017;135:201-203. 291. Wolff G, Lin Y, Karathanos A et al. Implantable cardioverter/defibrillators for primary prevention in dilated cardiomyopathy post-DANISH: an updated meta-analysis and systematic review of randomized controlled trials. Clin Res Cardiol 2017;106:501-513. 292. Singh JP, Hall WJ, McNitt S et al. Factors influencing appropriate firing of the implanted defibrillator for ventricular tachycardia/fibrillation: findings from the Multicenter Automatic Defibrillator Implantation Trial II (MADIT-II). J Am Coll Cardiol 2005;46:1712-1720. 293. Stecker EC, Chugh SS. Prediction of sudden cardiac death: next steps in pursuit of effective methodology. J Interv Card Electrophysiol 2011;31:101-107. 294. Gorgels AP, Gijsbers C, de Vreede-Swagemakers J, Lousberg A, Wellens HJ. Out-of-hospital cardiac arrest--the relevance of heart failure. The Maastricht Circulatory Arrest Registry. Eur Heart J 2003;24:1204-1209.

367 Template by Friedman & Morgan 2014 Word Chapter 7: References

295. Chugh SS, Jui J, Gunson K et al. Current burden of sudden cardiac death: multiple source surveillance versus retrospective death certificate-based review in a large U.S. community. J Am Coll Cardiol 2004;44:1268-1275. 296. Stecker EC, Vickers C, Waltz J et al. Population-based analysis of sudden cardiac death with and without left ventricular systolic dysfunction: two-year findings from the Oregon Sudden Unexpected Death Study. J Am Coll Cardiol 2006;47:1161-1166. 297. Al-Khatib SM, Shaw LK, O'Connor C, Kong M, Califf RM. Incidence and predictors of sudden cardiac death in patients with diastolic heart failure. J Cardiovasc Electrophysiol 2007;18:1231-1235. 298. Gold MR, Ip JH, Costantini O et al. Role of microvolt T-wave alternans in assessment of arrhythmia vulnerability among patients with heart failure and systolic dysfunction: primary results from the T- wave alternans sudden cardiac death in heart failure trial substudy. Circulation 2008;118:2022-2028. 299. Goldberger JJ, Cain ME, Hohnloser SH et al. American Heart Association/american College of Cardiology Foundation/heart Rhythm Society scientific statement on noninvasive risk stratification techniques for identifying patients at risk for sudden cardiac death: a scientific statement from the American Heart Association Council on Clinical Cardiology Committee on Electrocardiography and Arrhythmias and Council on Epidemiology and Prevention. Heart Rhythm 2008;5:e1-21. 300. Cubbon RM, Witte KK, Kearney LC et al. Performance of 2014 NICE defibrillator implantation guidelines in heart failure risk stratification. Heart 2016;102:735-740. 301. Fishman GI, Chugh SS, Dimarco JP et al. Sudden cardiac death prediction and prevention: report from a National Heart, Lung, and Blood Institute and Heart Rhythm Society Workshop. Circulation 2010;122:2335-2348. 302. Kirchhof P, Sipido KR, Cowie MR et al. The continuum of personalized cardiovascular medicine: a position paper of the European Society of Cardiology. Eur Heart J 2014;35:3250-3257. 303. Kobayashi Y, Yang S, Nykamp K et al. Pathogenic variant burden in the ExAC database: an empirical approach to evaluating population data for clinical variant interpretation. Genome Medicine 2017;9:13. 304. Walsh R, Buchan R, Wilk A et al. Defining the genetic architecture of hypertrophic cardiomyopathy: re-evaluating the role of non-sarcomeric genes. Eur Heart J 2017;10.1093/eurheartj/ehw603. 305. Pua CJ, Bhalshankar J, Miao K et al. Development of a Comprehensive Sequencing Assay for Inherited Cardiac Condition Genes. J Cardiovasc Transl Res 2016;9:3-11. 306. Elliott P. Cardiomyopathy. Diagnosis and management of dilated cardiomyopathy. Heart 2000;84:106- 112. 307. Rapezzi C, Arbustini E, Caforio AL et al. Diagnostic work-up in cardiomyopathies: bridging the gap between clinical phenotypes and final diagnosis. A position statement from the ESC Working Group on Myocardial and Pericardial Diseases. Eur Heart J 2013;34:1448-1458. 308. Perez-Serra A, Toro R, Sarquella-Brugada G et al. Genetic basis of dilated cardiomyopathy. Int J Cardiol 2016;224:461-472. 309. Kimura A. Molecular genetics and pathogenesis of cardiomyopathy. J Hum Genet 2016;61:41-50. 310. Favalli V, Serio A, Grasso M, Arbustini E. Genetic causes of dilated cardiomyopathy. Heart 2016;102:2004-2014. 311. Nagueh SF, Zoghbi WA. Role of Imaging in the Evaluation of Patients at Risk for Sudden Cardiac Death: Genotype-Phenotype Intersection. JACC Cardiovasc Imaging 2015;8:828-845. 312. Moulik M, Vatta M, Witt SH et al. ANKRD1, the gene encoding cardiac ankyrin repeat protein, is a novel dilated cardiomyopathy gene. J Am Coll Cardiol 2009;54:325-333. 313. gnomAD. 2016. http://gnomad.broadinstitute.org. Accessed:2/5/17. 314. Haas J, Frese KS, Peil B et al. Atlas of the clinical genetics of human dilated cardiomyopathy. Eur Heart J 2015;36:1123-1135a. 315. Lee S, Abecasis Gonçalo R, Boehnke M, Lin X. Rare-Variant Association Analysis: Study Designs and Statistical Tests. American Journal of Human Genetics 2014;95:5-23. 316. Guo Michael H, Dauber A, Lippincott Margaret F et al. Determinants of Power in Gene-Based Burden Testing for Monogenic Disorders. The American Journal of Human Genetics 2016;99:527-539. 317. Meienberg J, Bruggmann R, Oexle K, Matyas G. Clinical sequencing: is WGS the better WES? Human Genetics 2016;135:359-362. 318. Chen YC, Liu T, Yu CH, Chiang TY, Hwang CC. Effects of GC bias in next-generation-sequencing data on de novo genome assembly. PLoS One 2013;8:e62856. 319. Mestroni L, Brun F, Spezzacatene A, Sinagra G, Taylor MRG. Genetic causes of dilated cardiomyopathy. Progress in Pediatric Cardiology 2014;37:13-18. 320. Robinson JT, Thorvaldsdottir H, Winckler W et al. Integrative genomics viewer. Nat Biotech 2011;29:24-26. 321. Thorvaldsdottir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Brief Bioinform 2013;14:178-192.

368 Template by Friedman & Morgan 2014 Word Chapter 7: References

322. Akinrinade O, Alastalo TP, Koskenvuo JW. Relevance of Truncating Titin Mutations in Dilated Cardiomyopathy. Clin Genet 2016;90:49-54. 323. Mestroni L, Maisch B, McKenna WJ et al. Guidelines for the study of familial dilated cardiomyopathies. European Heart Journal 1999;20:93-102. 324. Petersen SE, Selvanayagam JB, Wiesmann F et al. Left ventricular non-compaction: insights from cardiovascular magnetic resonance imaging. J Am Coll Cardiol 2005;46:101-105. 325. Friedrich MG, Marcotte F. Cardiac magnetic resonance assessment of myocarditis. Circ Cardiovasc Imaging 2013;6:833-839. 326. Arany Z, Elkayam U. Peripartum Cardiomyopathy. Circulation 2016;133:1397-1409. 327. Burkett EL, Hershberger RE. Clinical and genetic issues in familial dilated cardiomyopathy. J Am Coll Cardiol 2005;45:969-981. 328. Committee HoCSaT. Alcohol guidelines: Eleventh Report of Session 2010-12. UK Parliament, 2012. 329. Royal Colleges of Physicians P, and General Practitioners UK. Alcohol and the heart in perspective: Sensible limits reaffirmed, 1995. 330. European Heart Rhythm A, Heart Rhythm S, Zipes DP et al. ACC/AHA/ESC 2006 guidelines for management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: a report of the American College of Cardiology/American Heart Association Task Force and the European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Develop Guidelines for Management of Patients With Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death). J Am Coll Cardiol 2006;48:e247-346. 331. Zou J, Tran D, Baalbaki M et al. An internal promoter underlies the difference in disease severity between N- and C-terminal truncation mutations of Titin in zebrafish. Elife 2015;4:e09406. 332. Guzzo-Merello G, Cobo-Marcos M, Gallego-Delgado M, Garcia-Pavia P. Alcoholic cardiomyopathy. World Journal of Cardiology 2014;6:771-781. 333. Roerecke M, Kaczorowski J, Tobe SW et al. The effect of a reduction in alcohol consumption on blood pressure: a systematic review and meta-analysis. The Lancet Public Health 2017;2:e108-e120. 334. Xin X, He J, Frontini MG et al. Effects of alcohol reduction on blood pressure: a meta-analysis of randomized controlled trials. Hypertension 2001;38:1112-1117. 335. Shimizu I, Minamino T. Physiological and pathological cardiac hypertrophy. J Mol Cell Cardiol 2016;97:245-262. 336. Xu L, Brink M. mTOR, cardiomyocytes and inflammation in cardiac hypertrophy. Biochim Biophys Acta 2016;1863:1894-1903. 337. Saxton RA, Sabatini DM. mTOR Signaling in Growth, Metabolism, and Disease. Cell 2017;168:960- 976. 338. Yano T, Shimoshige S, Miki T et al. Clinical impact of myocardial mTORC1 activation in nonischemic dilated cardiomyopathy. J Mol Cell Cardiol 2016;91:6-9. 339. Zhang D, Contu R, Latronico MVG et al. MTORC1 regulates cardiac function and myocyte survival through 4E-BP1 inhibition in mice. The Journal of Clinical Investigation 2010;120:2805-2816. 340. Hinson JT, Chopra A, Nafissi N et al. HEART DISEASE. Titin mutations in iPS cells define sarcomere insufficiency as a cause of dilated cardiomyopathy. Science 2015;349:982-986. 341. Shih YH, Dvornikov AV, Zhu P et al. Exon- and contraction-dependent functions of titin in sarcomere assembly. Development 2016;143:4713-4722. 342. Urbano-Marquez A, Estruch R, Navarro-Lopez F et al. The effects of alcoholism on skeletal and cardiac muscle. N Engl J Med 1989;320:409-415. 343. Fernandez-Sola J, Estruch R, Grau JM et al. The relation of alcoholic myopathy to cardiomyopathy. Ann Intern Med 1994;120:529-536. 344. Rodriguez A, Chawla K, Umoh NA et al. Alcohol and Apoptosis: Friends or Foes? Biomolecules 2015;5:3193-3203. 345. Mouton AJ, Ninh VK, El Hajj EC et al. Exposure to chronic alcohol accelerates development of wall stress and eccentric remodeling in rats with volume overload. J Mol Cell Cardiol 2016;97:15-23. 346. Yancy CW, Jessup M, Bozkurt B et al. 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2013;62:e147-239. 347. Arbustini E, Disertori M, Narula J. Primary Prevention of Sudden Arrhythmic Death in Dilated Cardiomyopathy: Current Guidelines and Risk Stratification. JACC Heart Fail 2017;5:39-43. 348. Metra M, Nodari S, Parrinello G et al. Marked improvement in left ventricular ejection fraction during long-term beta-blockade in patients with chronic heart failure: clinical correlates and prognostic significance. Am Heart J 2003;145:292-299. 349. Matsumura Y, Hoshikawa-Nagai E, Kubo T et al. Prediction of left ventricular reverse remodeling after therapy with angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers and β

369 Template by Friedman & Morgan 2014 Word Chapter 7: References

blockers in patients with idiopathic dilated cardiomyopathy. Cardiovascular Ultrasound 2015;13:14. 350. Kawai K, Takaoka H, Hata K, Yokota Y, Yokoyama M. Prevalence, predictors, and prognosis of reversal of maladaptive remodeling with intensive medical therapy in idiopathic dilated cardiomyopathy. Am J Cardiol 1999;84:671-676. 351. Arad M, Nussbaum T, Blechman I et al. Prevalence and clinical predictors of reverse remodeling in patients with dilated cardiomyopathy. Isr Med Assoc J 2014;16:405-411. 352. Pratali L, Picano E, Otasevic P et al. Prognostic significance of the dobutamine echocardiography test in idiopathic dilated cardiomyopathy. Am J Cardiol 2001;88:1374-1378. 353. Gorcsan J, 3rd, Murali S, Counihan PJ, Mandarino WA, Kormos RL. Right ventricular performance and contractile reserve in patients with severe heart failure. Assessment by pressure-area relations and association with outcome. Circulation 1996;94:3190-3197. 354. Di Salvo TG, Mathier M, Semigran MJ, Dec GW. Preserved right ventricular ejection fraction predicts exercise capacity and survival in advanced heart failure. J Am Coll Cardiol 1995;25:1143-1153. 355. Otasevic P, Popovic Z, Pratali L et al. Right vs. left ventricular contractile reserve in one-year prognosis of patients with idiopathic dilated cardiomyopathy: assessment by dobutamine stress echocardiography. Eur J Echocardiogr 2005;6:429-434. 356. Lancellotti P, Senechal M, Moonen M et al. Myocardial contractile reserve during exercise predicts left ventricular reverse remodelling after cardiac resynchronization therapy. Eur J Echocardiogr 2009;10:663-668. 357. Neskovic AN, Otasevic P. Stress-echocardiography in idiopathic dilated cardiomyopathy: instructions for use. Cardiovasc Ultrasound 2005;3:3. 358. Aletras AH, Ding S, Balaban RS, Wen H. DENSE: displacement encoding with stimulated echoes in cardiac functional MRI. J Magn Reson 1999;137:247-252. 359. Wedeen VJ, Weisskoff RM, Reese TG et al. Motionless movies of myocardial strain-rates using stimulated echoes. Magn Reson Med 1995;33:401-408. 360. Ohshima S, Isobe S, Izawa H et al. Cardiac sympathetic dysfunction correlates with abnormal myocardial contractile reserve in dilated cardiomyopathy patients. J Am Coll Cardiol 2005;46:2061- 2068. 361. Bargiggia GS, Bertucci C, Recusani F et al. A new method for estimating left ventricular dP/dt by continuous wave Doppler-echocardiography. Validation studies at cardiac catheterization. Circulation 1989;80:1287-1292. 362. Kobayashi M, Izawa H, Cheng XW et al. Dobutamine stress testing as a diagnostic tool for evaluation of myocardial contractile reserve in asymptomatic or mildly symptomatic patients with dilated cardiomyopathy. JACC Cardiovasc Imaging 2008;1:718-726. 363. Tio RA, Slart R, de Boer RA et al. Reduced regional myocardial perfusion reserve is associated with impaired contractile performance in idiopathic dilated cardiomyopathy. Netherlands Heart Journal 2009;17:470-474. 364. Skalidis EI, Parthenakis FI, Patrianakos AP, Hamilos MI, Vardas PE. Regional coronary flow and contractile reserve in patients with idiopathic dilated cardiomyopathy. Journal of the American College of Cardiology 2004;44:2027-2032. 365. Otasevic P, Popovic ZB, Vasiljevic JD et al. Relation of myocardial histomorphometric features and left ventricular contractile reserve assessed by high-dose dobutamine stress echocardiography in patients with idiopathic dilated cardiomyopathy. Eur J Heart Fail 2005;7:49-56. 366. Xue H, Shah S, Greiser A et al. Motion correction for myocardial T1 mapping using image registration with synthetic image estimation. Magn Reson Med 2012;67:1644-1655. 367. Moon JC, Messroghli DR, Kellman P et al. Myocardial T1 mapping and extracellular volume quantification: a Society for Cardiovascular Magnetic Resonance (SCMR) and CMR Working Group of the European Society of Cardiology consensus statement. Journal of Cardiovascular Magnetic Resonance 2013;15:92. 368. Spottiswoode BS, Zhong X, Hess AT et al. Tracking myocardial motion from cine DENSE images using spatiotemporal phase unwrapping and temporal fitting. IEEE Trans Med Imaging 2007;26:15-30. 369. Gilliam AD, Suever JD. DENSE analysis: Cine DENSE Processing Software. 2016. https://github.com/denseanalysis/denseanalysis. Accessed:26.7.17. 370. Scott AD, Tayal U, Nielles-Vallespin S et al. Accelerating cine DENSE using a zonal excitation. Journal of Cardiovascular Magnetic Resonance 2016;18:O50. 371. Simonetti OP, Kim RJ, Fieno DS et al. An improved MR imaging technique for the visualization of myocardial infarction. Radiology 2001;218:215-223. 372. Eichhorn EJ, Grayburn PA, Mayer SA et al. Myocardial contractile reserve by dobutamine stress echocardiography predicts improvement in ejection fraction with beta-blockade in patients with heart failure: the Beta-Blocker Evaluation of Survival Trial (BEST). Circulation 2003;108:2336-2341.

370 Template by Friedman & Morgan 2014 Word Chapter 7: References

373. Barnett AG, van der Pols JC, Dobson AJ. Regression to the mean: what it is and how to deal with it. Int J Epidemiol 2005;34:215-220. 374. Textor J, Hardt J, Knuppel S. DAGitty: a graphical tool for analyzing causal diagrams. Epidemiology 2011;22:745. 375. Scott AE, Semple SI, Redpath TW, Hillis GS. Low-dose dobutamine adds incremental value to late gadolinium enhancement cardiac magnetic resonance in the prediction of adverse remodelling following acute myocardial infarction. Eur Heart J Cardiovasc Imaging 2013;14:906-913. 376. Drakos SG, Pagani FD, Lundberg MS, Baldwin JT. Advancing the Science of Myocardial Recovery With Mechanical Circulatory Support: A Working Group of the National, Heart, Lung and Blood Institute. J Card Fail 2017;23:416-421. 377. Seghatol FF, Shah DJ, Diluzio S et al. Relation between contractile reserve and improvement in left ventricular function with beta-blocker therapy in patients with heart failure secondary to ischemic or idiopathic dilated cardiomyopathy. Am J Cardiol 2004;93:854-859. 378. Kehat I, Molkentin JD. Molecular pathways underlying cardiac remodeling during pathophysiological stimulation. Circulation 2010;122:2727-2735. 379. Bello D, Shah DJ, Farah GM et al. Gadolinium cardiovascular magnetic resonance predicts reversible myocardial dysfunction and remodeling in patients with heart failure undergoing beta-blocker therapy. Circulation 2003;108:1945-1953. 380. Hutchinson KR, Stewart JA, Jr., Lucchesi PA. Extracellular matrix remodeling during the progression of volume overload-induced heart failure. J Mol Cell Cardiol 2010;48:564-569. 381. Buckberg G, Hoffman JI, Mahajan A, Saleh S, Coghlan C. Cardiac mechanics revisited: the relationship of cardiac architecture to ventricular function. Circulation 2008;118:2571-2587. 382. Sampaio F, Lamata P, Bettencourt N et al. Assessment of cardiovascular physiology using dobutamine stress cardiovascular magnetic resonance reveals impaired contractile reserve in patients with cirrhotic cardiomyopathy. J Cardiovasc Magn Reson 2015;17:61. 383. Chimura M, Onishi T, Tsukishiro Y et al. Longitudinal strain combined with delayed-enhancement magnetic resonance improves risk stratification in patients with dilated cardiomyopathy. Heart 2017;103:679-686. 384. Ersboll M, Valeur N, Mogensen UM et al. Prediction of all-cause mortality and heart failure admissions from global left ventricular longitudinal strain in patients with acute myocardial infarction and preserved left ventricular ejection fraction. J Am Coll Cardiol 2013;61:2365-2373. 385. Mignot A, Donal E, Zaroui A et al. Global longitudinal strain as a major predictor of cardiac events in patients with depressed left ventricular function: a multicenter study. J Am Soc Echocardiogr 2010;23:1019-1024. 386. Adamo L, Perry A, Novak E et al. Abnormal Global Longitudinal Strain Predicts Future Deterioration of Left Ventricular Function in Heart Failure Patients With a Recovered Left Ventricular Ejection Fraction. Circ Heart Fail 2017;10. 387. Mangion K, Gao H, McComb C et al. A Novel Method for Estimating Myocardial Strain: Assessment of Deformation Tracking Against Reference Magnetic Resonance Methods in Healthy Volunteers. Sci Rep 2016;6:38774. 388. Wilcox JE, Yancy CW. Heart Failure-A New Phenotype Emerges. JAMA Cardiol 2016;1:507-509. 389. Ambardekar AV, Buttrick PM. Reverse remodeling with left ventricular assist devices: a review of clinical, cellular, and molecular effects. Circ Heart Fail 2011;4:224-233. 390. Goldenberg I, Moss AJ, Hall WJ et al. Predictors of response to cardiac resynchronization therapy in the Multicenter Automatic Defibrillator Implantation Trial with Cardiac Resynchronization Therapy (MADIT-CRT). Circulation 2011;124:1527-1536. 391. Rickard J, Michtalik H, Sharma R et al. Predictors of response to cardiac resynchronization therapy: A systematic review. Int J Cardiol 2016;225:345-352. 392. Chen Z, Sohal M, Sammut E et al. Focal but not diffuse myocardial fibrosis burden quantification using cardiac magnetic resonance imaging predicts left ventricular reverse modeling following cardiac resynchronization therapy. J Cardiovasc Electrophysiol 2015;27:203-209. 393. Hsu JC, Solomon SD, Bourgoun M et al. Predictors of super-response to cardiac resynchronization therapy and associated improvement in clinical outcome: the MADIT-CRT (multicenter automatic defibrillator implantation trial with cardiac resynchronization therapy) study. J Am Coll Cardiol 2012;59:2366-2373. 394. Aleksova A, Merlo M, Zecchin M et al. Impact of atrial fibrillation on outcome of patients with idiopathic dilated cardiomyopathy: data from the Heart Muscle Disease Registry of Trieste. Clin Med Res 2010;8:142-149. 395. Mehrotra P, Labib SB, Schick EC. Differential effects of dobutamine versus treadmill exercise on left ventricular volume and wall stress. J Am Soc Echocardiogr 2012;25:911-918.

371 Template by Friedman & Morgan 2014 Word Chapter 7: References

396. Konstam MA, Abboud FM. Ejection Fraction: Misunderstood and Overrated (Changing the Paradigm in Categorizing Heart Failure). Circulation 2017;135:717-719. 397. Kalogeropoulos AP, Fonarow GC, Georgiopoulou V et al. Characteristics and Outcomes of Adult Outpatients With Heart Failure and Improved or Recovered Ejection Fraction. JAMA Cardiol 2016;1:510-518. 398. Ruiz-Zamora I, Rodriguez-Capitan J, Guerrero-Molina A et al. Incidence and prognosis implications of long term left ventricular reverse remodeling in patients with dilated cardiomyopathy. Int J Cardiol 2015;203:1114-1121. 399. Cho JY, Kim KH, Song JE et al. Predictors of Left Ventricular Functional Recovery and Their Impact on Clinical Outcomes in Patients With Newly Diagnosed Dilated Cardiomyopathy and Heart Failure. Heart Lung Circ 2017;10.1016/j.hlc.2017.02.013. 400. van der Heijden AC, Hoke U, Thijssen J et al. Super-responders to cardiac resynchronization therapy remain at risk for ventricular arrhythmias and benefit from defibrillator treatment. Eur J Heart Fail 2014;16:1104-1111. 401. Hicks KA, Tcheng JE, Bozkurt B et al. 2014 ACC/AHA Key Data Elements and Definitions for Cardiovascular Endpoint Events in Clinical Trials: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Data Standards (Writing Committee to Develop Cardiovascular Endpoints Data Standards). Circulation 2015;132:302-361. 402. Altman DG, Andersen PK. Bootstrap investigation of the stability of a Cox regression model. Stat Med 1989;8:771-783. 403. Walter S, Tiemeier H. Variable selection: current practice in epidemiological studies. Eur J Epidemiol 2009;24:733-736. 404. Hernan MA, Hernandez-Diaz S, Werler MM, Mitchell AA. Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. Am J Epidemiol 2002;155:176- 184. 405. Greenland S. Invited commentary: variable selection versus shrinkage in the control of multiple confounders. Am J Epidemiol 2008;167:523-529. 406. Hosmer Jr DW, Lemeshow S, Sturdivant RX. Applied logistic regression: John Wiley & Sons, 2013. 407. Hosmer DJ. S.. Lemeshow;(1999).‘Applied Survival Analysis. Regression Modeling of Time to Event Data.’. Wiley: New York. 408. Gold MR, Daubert C, Abraham WT et al. The effect of reverse remodeling on long-term survival in mildly symptomatic patients with heart failure receiving cardiac resynchronization therapy: results of the REVERSE study. Heart Rhythm 2015;12:524-530. 409. McManus DD, Shah SJ, Fabi MR et al. Prognostic value of left ventricular end-systolic volume index as a predictor of heart failure hospitalization in stable coronary artery disease: data from the Heart and Soul Study. J Am Soc Echocardiogr 2009;22:190-197. 410. Pontone G, Guaricci AI, Andreini D et al. Prognostic Benefit of Cardiac Magnetic Resonance Over Transthoracic Echocardiography for the Assessment of Ischemic and Nonischemic Dilated Cardiomyopathy Patients Referred for the Evaluation of Primary Prevention Implantable Cardioverter- Defibrillator Therapy. Circ Cardiovasc Imaging 2016;9: 004956. 411. Wijesurendra RS, Rider OJ, Neubauer S. Left Atrial Volumes in Health and Disease Measured Using Cardiac Magnetic Resonance. Circ Cardiovasc Imaging 2017;10. 412. Zemrak F, Ambale-Venkatesh B, Captur G et al. Left Atrial Structure in Relationship to Age, Sex, Ethnicity, and Cardiovascular Risk Factors: MESA (Multi-Ethnic Study of Atherosclerosis). Circ Cardiovasc Imaging 2017;10. 413. Shen L, Jhund PS, Petrie MC et al. Declining Risk of Sudden Death in Heart Failure. N Engl J Med 2017;377:41-51. 414. Peduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events per independent variable in proportional hazards regression analysis. II. Accuracy and precision of regression estimates. J Clin Epidemiol 1995;48:1503-1510. 415. Shah SJ, Katz DH, Selvaraj S et al. Phenomapping for novel classification of heart failure with preserved ejection fraction. Circulation 2015;131:269-279. 416. Zipes DP, Camm AJ, Borggrefe M et al. ACC/AHA/ESC 2006 Guidelines for Management of Patients With Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death: a report of the American College of Cardiology/American Heart Association Task Force and the European Society of Cardiology Committee for Practice Guidelines (writing committee to develop Guidelines for Management of Patients With Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death): developed in collaboration with the European Heart Rhythm Association and the Heart Rhythm Society. Circulation 2006;114:e385-484.

372 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

8 APPENDICES

8.1 Gene panels Appendix Table 1: Genes sequenced on Tru Sight Cardio Gene Panel- 174 genes

Gene Gene Name ICC Diseases

LQTS, JLNS, KCNQ1 potassium voltage-gated channel, KQT-like subfamily, member 1 SQTS, HCM, AF

LQTS, SQTS, KCNH2 potassium voltage-gated channel, subfamily H (eag-related), member 2 BrS

LQTS, BrS, SCN5A sodium channel, voltage-gated, type V, alpha subunit DCM, ARVD/C, AF

ANK2 ankyrin 2, neuronal LQTS

LQTS, JLNS, KCNE1 potassium voltage-gated channel, Isk-related family, member 1 CPVT, AF

KCNE2 potassium voltage-gated channel, Isk-related family, member 2 LQTS, AF

LQTS, SQTS, KCNJ2 potassium inwardly-rectifying channel, subfamily J, member 2 CPVT, AF

CACNA1C calcium channel, voltage-dependent, L type, alpha 1C subunit LQTS, BrS, HCM

CAV3 caveolin 3 LQTS, HCM

SCN4B sodium channel, voltage-gated, type IV, beta LQTS, AF

AKAP9 A kinase (PRKA) anchor protein (yotiao) 9 LQTS

SNTA1 syntrophin, alpha 1 (dystrophin-associated protein A1, 59kDa, acidic component) LQTS

KCNJ5 potassium inwardly-rectifying channel, subfamily J, member 5 LQTS

ABCC9 ATP-binding cassette, sub-family C (CFTR/MRP), member 9 BrS, DCM, AF

TNNC1 troponin C type 1 (slow) HCM, DCM

ANKRD1 ankyrin repeat domain 1 (cardiac muscle) HCM, DCM

HCM, DCM, LMNA lamin A/C ARVD/C, AF

HCM, DCM, DES desmin ARVD/C, RCM

RBM20 RNA binding motif protein 20 DCM

VCL vinculin HCM, DCM

HCM, DCM, ACTN2 , alpha 2 ncCM

373 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

HCM, DCM, MYH7 myosin, heavy chain 7, cardiac muscle, beta RCM, ncCM

HCM, DCM, TNNT2 troponin T type 2 (cardiac) RCM, ncCM

LDB3 LIM domain binding 3 HCM, DCM

HCM, DCM, CSRP3 cysteine and glycine-rich protein 3 (cardiac LIM protein) ncCM

HCM, DCM, TNNI3 troponin I type 3 (cardiac) RCM

HCM, DCM, MYBPC3 myosin binding protein C, cardiac ncCM

HCM, DCM, TPM1 tropomyosin 1 (alpha) RCM, ncCM

HCM, DCM, ACTC1 actin, alpha, cardiac muscle 1 RCM, ncCM

TCAP titin-cap (telethonin) HCM, DCM

MYH6 myosin, heavy chain 6, cardiac muscle, alpha HCM, DCM, AF

HCM, DCM, PLN phospholamban ARVD/C

HCM, DCM, TTN titin ARVD/C

MYLK2 myosin light chain kinase 2 HCM, FAA

MYL2 myosin, light chain 2, regulatory, cardiac, slow HCM, RCM

CALR3 3 HCM

JPH2 junctophilin 2 HCM, AF

MYL3 myosin, light chain 3, alkali; ventricular, skeletal, slow HCM, RCM

MYOZ2 myozenin 2 HCM

DSG2 desmoglein 2 DCM, ARVD/C

BrS, DCM, PKP2 plakophilin 2 ARVD/C

DSP desmoplakin DCM, ARVD/C

TGFB3 transforming growth factor, beta 3 ARVD/C, LDS

DCM, ARVD/C, DSC2 desmocollin 2 AF

TMEM43 transmembrane protein 43 ARVD/C

JUP junction plakoglobin DCM, ARVD/C

374 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

LQTS, CPVT, RYR2 2 (cardiac) ARVD/C

ACTA1 actin, alpha 1, skeletal muscle HCM, DCM

CPVT, HCM, CASQ2 2 (cardiac muscle) DCM, ncCM

COX15 COX15 homolog, cytochrome c oxidase assembly protein (yeast) HCM

CRYAB crystallin, alpha B HCM, DCM

CTF1 cardiotrophin 1 DCM

FKTN fukutin DCM

FXN frataxin HCM

ILK integrin-linked kinase DCM

LAMA2 laminin, alpha 2 DCM

LAMA4 laminin, alpha 4 DCM

MYO6 myosin VI HCM

HCM, DCM, MYPN myopalladin RCM

NEXN nexilin (F actin binding protein) HCM, DCM

RAF1 v-raf-1 murine leukemia viral oncogene homolog 1 HCM, DCM, NS

solute carrier family 25 (mitochondrial carrier; adenine nucleotide translocator), SLC25A4 HCM member 4

TTR transthyretin

TXNRD2 thioredoxin reductase 2 DCM

DMD dystrophin DCM

PRKAG2 protein kinase, AMP-activated, gamma 2 non-catalytic subunit HCM

SGCD sarcoglycan, delta (35kDa dystrophin-associated glycoprotein) DCM

TAZ tafazzin DCM, ncCM

TMPO thymopoietin DCM

EYA4 eyes absent homolog 4 (Drosophila) DCM

CACNA2D1 calcium channel, voltage-dependent, alpha 2/delta subunit 1 SQTS, BrS

CACNB2 calcium channel, voltage-dependent, beta 2 subunit BrS

GPD1L glycerol-3-phosphate dehydrogenase 1-like BrS

KCND3 potassium voltage-gated channel, Shal-related subfamily, member 3 BrS, AF

375 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

KCNE3 potassium voltage-gated channel, Isk-related family, member 3 LQTS, BrS, AF

SCN1B sodium channel, voltage-gated, type I, beta BrS

SCN3B sodium channel, voltage-gated, type III, beta BrS, AF

HCN4 hyperpolarization activated cyclic nucleotide-gated potassium channel 4 BrS, ncCM, AF

GLA galactosidase, alpha HCM

SOS1 son of sevenless homolog 1 (Drosophila) HCM, NS

PTPN11 protein tyrosine phosphatase, non-receptor type 11 HCM, NS

KRAS v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog NS

SDHA succinate dehydrogenase complex, subunit A, flavoprotein (Fp) DCM

BAG3 BCL2-associated athanogene 3 DCM

GATAD1 GATA zinc finger domain containing 1 DCM

ALMS1 Alstrom syndrome 1 DCM

DNAJC19 DnaJ (Hsp40) homolog, subfamily C, member 19 DCM

DTNA dystrobrevin, alpha DCM, ncCM

EMD emerin DCM, AF

FHL2 four and a half LIM domains 2 DCM

FKRP fukutin related protein DCM

hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase HADHA DCM (trifunctional protein), alpha subunit

HFE hemochromatosis DCM

LAMP2 lysosomal-associated 2 HCM, DCM

PDLIM3 PDZ and LIM domain 3 HCM, DCM

SGCB sarcoglycan, beta (43kDa dystrophin-associated glycoprotein) DCM

SGCG sarcoglycan, gamma (35kDa dystrophin-associated glycoprotein) DCM

TBX20 T-box 20 DCM

NPPA natriuretic peptide A DCM, AF

KCNA5 potassium voltage-gated channel, shaker-related subfamily, member 5 AF

GJA5 gap junction protein, alpha 5, 40kDa AF

ACTA2 actin, alpha 2, smooth muscle, aorta FAA

COL3A1 collagen, type III, alpha 1 FAA

376 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

KCNJ8 potassium inwardly-rectifying channel, subfamily J, member 8 BrS, AF

HSPB8 heat shock 22kDa protein 8

MYH11 myosin, heavy chain 11, smooth muscle FAA

SMAD3 SMAD family member 3 LDS, FAA

LMF1 lipase maturation factor 1 FHT

TGFBR2 transforming growth factor, beta receptor II (70/80kDa) MFS, LDS, FAA

TRIM63 tripartite motif containing 63, E3 ubiquitin protein ligase HCM

DPP6 dipeptidyl-peptidase 6

TGFBR1 transforming growth factor, beta receptor 1 MFS, LDS, FAA

APOA5 apolipoprotein A-V FHT, FHL

MYLK myosin light chain kinase FAA

ZBTB17 zinc finger and BTB domain containing 17

LDLR low density lipoprotein receptor FH

LPL lipoprotein lipase FHT, FHL

RANGRF RAN guanine nucleotide release factor BrS

MFS, LDS, FAA, FBN1 1 AVD

CALM1 1 (phosphorylase kinase, delta) LQTS, CPVT

GAA glucosidase, alpha; acid HCM

RYR1 ryanodine receptor 1 (skeletal)

BRAF v-raf murine sarcoma viral oncogene homolog B1 HCM, NS

DOLK dolichol kinase DCM

FHL1 four and a half LIM domains 1 HCM

KLF10 Kruppel-like factor 10 HCM

MURC muscle-related coiled-coil protein DCM

NRAS neuroblastoma RAS viral (v-ras) oncogene homolog NS

TRDN triadin CPVT

APOB apolipoprotein B (including Ag(x) antigen) FH, FHT

PCSK9 proprotein convertase subtilisin/kexin type 9 FH

GCKR glucokinase (hexokinase 4) regulator FH, FHT

377 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

ELN elastin FAA, AVD

APOE apolipoprotein E FH, FHT, FHL

GPIHBP1 glycosylphosphatidylinositol anchored high density lipoprotein binding protein 1 FHT, FHL

NOTCH1 notch 1 FAA, AVD

APOC2 apolipoprotein C-II FHT, FHL

ZHX3 zinc fingers and homeoboxes 3 FHT

CETP cholesteryl ester transfer protein, plasma FH, FHL

TGFB2 transforming growth factor, beta 2 LDS, FAA

CREB3L3 cAMP responsive element binding protein 3-like 3 FHT

LDLRAP1 low density lipoprotein receptor adaptor protein 1 FH

APOA4 apolipoprotein A-IV FHL

SREBF2 sterol regulatory element binding transcription factor 2 FH

LTBP2 latent transforming growth factor beta binding protein 2 MFS

ABCG5 ATP-binding cassette, sub-family G (WHITE), member 5

ABCG8 ATP-binding cassette, sub-family G (WHITE), member 8 FH

PRDM16 PR domain containing 16 DCM, ncCM

NKX2-5 NK2 homeobox 5 AF

TBX5 T-box 5

TBX3 T-box 3

SALL4 sal-like 4 (Drosophila)

SEPN1 selenoprotein N, 1

SLC2A10 solute carrier family 2 (facilitated glucose transporter), member 10 FAA

MIB1 mindbomb E3 ubiquitin protein ligase 1 ncCM

MAP2K1 mitogen-activated protein kinase kinase 1 HCM, NS

MAP2K2 mitogen-activated protein kinase kinase 2 HCM, NS

SHOC2 soc-2 suppressor of clear homolog (C. elegans) NS

JAG1 jagged 1

ZIC3 Zic family member 3

PRKAR1A protein kinase, cAMP-dependent, regulatory, type I, alpha

CRELD1 cysteine-rich with EGF-like domains 1

378 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

NODAL nodal homolog (mouse)

EFEMP2 EGF containing -like extracellular matrix protein 2 FAA

CBS cystathionine-beta-synthase

SMAD4 SMAD family member 4 FAA

FBN2 fibrillin 2 FAA

HRAS v-Ha-ras Harvey rat sarcoma viral oncogene homolog NS

COL5A1 collagen, type V, alpha 1

COL5A2 collagen, type V, alpha 2

CBL Cbl proto-oncogene, E3 ubiquitin protein ligase NS

SCN2B sodium channel, voltage-gated, type II, beta subunit BrS, AF

TRPM4 transient receptor potential cation channel, subfamily M, member 4 BrS

SCO2 SCO2 cytochrome c oxidase assembly protein

379 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Appendix Table 2: Genes sequenced on Inherited Cardiac Conditions (ICC) Gene Panels Versions 4-6– 201 genes

Gene Gene Name ICC Diseases

LQTS, JLNS, SQTS, KCNQ1 potassium voltage-gated channel, KQT-like subfamily, member 1 HCM, AF

KCNH2 potassium voltage-gated channel, subfamily H (eag-related), member 2 LQTS, SQTS, BrS

LQTS, BrS, DCM, SCN5A sodium channel, voltage-gated, type V, alpha subunit ARVD/C, AF

ANK2 ankyrin 2, neuronal LQTS

KCNE1 potassium voltage-gated channel, Isk-related family, member 1 LQTS, JLNS, CPVT, AF

KCNE2 potassium voltage-gated channel, Isk-related family, member 2 LQTS, AF

LQTS, SQTS, CPVT, KCNJ2 potassium inwardly-rectifying channel, subfamily J, member 2 AF

CACNA1C calcium channel, voltage-dependent, L type, alpha 1C subunit LQTS, BrS, HCM

CAV3 caveolin 3 LQTS, HCM

SCN4B sodium channel, voltage-gated, type IV, beta LQTS, AF

AKAP9 A kinase (PRKA) anchor protein (yotiao) 9 LQTS

SNTA1 syntrophin, alpha 1 (dystrophin-associated protein A1, 59kDa, acidic component) LQTS

KCNJ5 potassium inwardly-rectifying channel, subfamily J, member 5 LQTS

ABCC9 ATP-binding cassette, sub-family C (CFTR/MRP), member 9 BrS, DCM, AF

PSEN1 presenilin 1 DCM

TNNC1 troponin C type 1 (slow) HCM, DCM

PSEN2 presenilin 2 (Alzheimer disease 4) DCM

ANKRD1 ankyrin repeat domain 1 (cardiac muscle) HCM, DCM

HCM, DCM, ARVD/C, LMNA lamin A/C AF

HCM, DCM, ARVD/C, DES desmin RCM

RBM20 RNA binding motif protein 20 DCM

VCL vinculin HCM, DCM

ACTN2 actinin, alpha 2 HCM, DCM, ncCM

HCM, DCM, RCM, MYH7 myosin, heavy chain 7, cardiac muscle, beta ncCM

HCM, DCM, RCM, TNNT2 troponin T type 2 (cardiac) ncCM

380 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

LDB3 LIM domain binding 3 HCM, DCM

CSRP3 cysteine and glycine-rich protein 3 (cardiac LIM protein) HCM, DCM, ncCM

TNNI3 troponin I type 3 (cardiac) HCM, DCM, RCM

MYBPC3 myosin binding protein C, cardiac HCM, DCM, ncCM

HCM, DCM, RCM, TPM1 tropomyosin 1 (alpha) ncCM

HCM, DCM, RCM, ACTC1 actin, alpha, cardiac muscle 1 ncCM

TCAP titin-cap (telethonin) HCM, DCM

MYH6 myosin, heavy chain 6, cardiac muscle, alpha HCM, DCM, AF

PLN phospholamban HCM, DCM, ARVD/C

TTN titin HCM, DCM, ARVD/C

MYLK2 myosin light chain kinase 2 HCM, FAA

MYL2 myosin, light chain 2, regulatory, cardiac, slow HCM, RCM

CALR3 calreticulin 3 HCM

JPH2 junctophilin 2 HCM, AF

MYL3 myosin, light chain 3, alkali; ventricular, skeletal, slow HCM, RCM

MYOZ2 myozenin 2 HCM

DSG2 desmoglein 2 DCM, ARVD/C

PKP2 plakophilin 2 BrS, DCM, ARVD/C

DSP desmoplakin DCM, ARVD/C

TGFB3 transforming growth factor, beta 3 ARVD/C, LDS

DSC2 desmocollin 2 DCM, ARVD/C, AF

TMEM43 transmembrane protein 43 ARVD/C

JUP junction plakoglobin DCM, ARVD/C

RYR2 ryanodine receptor 2 (cardiac) LQTS, CPVT, ARVD/C

CPVT, HCM, DCM, CASQ2 calsequestrin 2 (cardiac muscle) ncCM

CRYAB crystallin, alpha B HCM, DCM

CTF1 cardiotrophin 1 DCM

FKTN fukutin DCM

FLT1 fms-related tyrosine kinase 1 (vascular endothelial growth factor/vascular DCM

381 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

permeability factor receptor)

FOXD4 forkhead box D4 DCM

FXN frataxin HCM

ILK integrin-linked kinase DCM

LAMA2 laminin, alpha 2 DCM

LAMA4 laminin, alpha 4 DCM

MYPN myopalladin HCM, DCM, RCM

NEXN nexilin (F actin binding protein) HCM, DCM

RAF1 v-raf-1 murine leukemia viral oncogene homolog 1 HCM, DCM, NS

SGCA sarcoglycan, alpha (50kDa dystrophin-associated glycoprotein)

SYNE1 repeat containing, nuclear envelope 1 DCM

SYNM , protein DCM

TTR transthyretin

DMD dystrophin DCM

PRKAG2 protein kinase, AMP-activated, gamma 2 non-catalytic subunit HCM

SGCD sarcoglycan, delta (35kDa dystrophin-associated glycoprotein) DCM

TAZ tafazzin DCM, ncCM

TMPO thymopoietin DCM

TP63 tumor protein p63 ARVD/C

EYA4 eyes absent homolog 4 (Drosophila) DCM

CACNA2D1 calcium channel, voltage-dependent, alpha 2/delta subunit 1 SQTS, BrS

CACNB2 calcium channel, voltage-dependent, beta 2 subunit BrS

GPD1L glycerol-3-phosphate dehydrogenase 1-like BrS

KCNE3 potassium voltage-gated channel, Isk-related family, member 3 LQTS, BrS, AF

SCN1B sodium channel, voltage-gated, type I, beta BrS

SCN3B sodium channel, voltage-gated, type III, beta BrS, AF

HCN4 hyperpolarization activated cyclic nucleotide-gated potassium channel 4 BrS, ncCM, AF

GLA galactosidase, alpha HCM

SOS1 son of sevenless homolog 1 (Drosophila) HCM, NS

PTPN11 protein tyrosine phosphatase, non-receptor type 11 HCM, NS

382 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

KRAS v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog NS

SDHA succinate dehydrogenase complex, subunit A, flavoprotein (Fp) DCM

BAG3 BCL2-associated athanogene 3 DCM

ALMS1 Alstrom syndrome 1 DCM

APOA1 apolipoprotein A-I DCM, FHL

DNAJC19 DnaJ (Hsp40) homolog, subfamily C, member 19 DCM

DTNA dystrobrevin, alpha DCM, ncCM

EMD emerin DCM, AF

FHL2 four and a half LIM domains 2 DCM

FKRP fukutin related protein DCM

hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase HADHA DCM (trifunctional protein), alpha subunit

HFE hemochromatosis DCM

HOPX HOP homeobox DCM

LAMP2 lysosomal-associated membrane protein 2 HCM, DCM

NFKB1 nuclear factor of kappa light polypeptide gene enhancer in B-cells 1 DCM

PDLIM3 PDZ and LIM domain 3 HCM, DCM

PLEC DCM, FH

SGCB sarcoglycan, beta (43kDa dystrophin-associated glycoprotein) DCM

SGCG sarcoglycan, gamma (35kDa dystrophin-associated glycoprotein) DCM

SOD2 superoxide dismutase 2, mitochondrial DCM

TBX20 T-box 20 DCM

CALM3 calmodulin 3 (phosphorylase kinase, delta) CPVT

HSP90AB1 heat shock protein 90kDa alpha (cytosolic), class B member 1

NDRG4 NDRG family member 4

ATG5 autophagy related 5

UNC45B unc-45 homolog B (C. elegans)

CCT3 chaperonin containing TCP1, subunit 3 (gamma)

CNOT3 CCR4-NOT transcription complex, subunit 3

LSM14A LSM14A, SCD6 homolog A (S. cerevisiae)

383 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

CALM2 calmodulin 2 (phosphorylase kinase, delta) LQTS

RBM7 RNA binding motif protein 7

RBM4 RNA binding motif protein 4

RXRA retinoid X receptor, alpha

NR1H2 nuclear receptor subfamily 1, group H, member 2

HSPB1 heat shock 27kDa protein 1

CCT4 chaperonin containing TCP1, subunit 4 (delta)

serpin peptidase inhibitor, clade E (nexin, plasminogen activator inhibitor type 1), SERPINE1 member 1

GPR183 G protein-coupled receptor 183

STUB1 STIP1 homology and U-box containing protein 1, E3 ubiquitin protein ligase

NR1H3 nuclear receptor subfamily 1, group H, member 3

UBE4B ubiquitination factor E4B

HSPB6 heat shock protein, alpha-crystallin-related, B6

CCT6B chaperonin containing TCP1, subunit 6B (zeta 2)

RXRG retinoid X receptor, gamma

AGL amylo-alpha-1, 6-glucosidase, 4-alpha-glucanotransferase

KCNJ8 potassium inwardly-rectifying channel, subfamily J, member 8 BrS, AF

CACNA1D calcium channel, voltage-dependent, L type, alpha 1D subunit

KCNQ2 potassium voltage-gated channel, KQT-like subfamily, member 2

CCT6A chaperonin containing TCP1, subunit 6A (zeta 1)

PFDN1 prefoldin subunit 1

SF3B3 splicing factor 3b, subunit 3, 130kDa

HSPB8 heat shock 22kDa protein 8

SLC8A1 solute carrier family 8 (sodium/calcium exchanger), member 1

SNRNP70 small nuclear ribonucleoprotein 70kDa (U1)

SCN10A sodium channel, voltage-gated, type X, alpha subunit BrS, AF

MBNL2 muscleblind-like splicing regulator 2

PRPF8 PRP8 pre-mRNA processing factor 8 homolog (S. cerevisiae)

CCT2 chaperonin containing TCP1, subunit 2 (beta)

384 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

RBM4B RNA binding motif protein 4B

RNF207 ring finger protein 207 LQTS

FOXO3 forkhead box O3

C21orf7 chromosome 21 open reading frame 7

CCT5 chaperonin containing TCP1, subunit 5 (epsilon)

VBP1 von Hippel-Lindau binding protein 1

HNRNPM heterogeneous nuclear ribonucleoprotein M

ADRB2 adrenoceptor beta 2, surface

HSPB7 heat shock 27kDa protein family, member 7 (cardiovascular)

ADRA2B adrenoceptor alpha 2B

RBM45 RNA binding motif protein 45

RBM39 RNA binding motif protein 39

TRIM55 tripartite motif containing 55 HCM

UBE2D3 ubiquitin-conjugating enzyme E2D 3

PTBP1 polypyrimidine tract binding protein 1

TRIM63 tripartite motif containing 63, E3 ubiquitin protein ligase HCM

MDM2 Mdm2, p53 E3 ubiquitin protein ligase homolog (mouse)

SNRPB small nuclear ribonucleoprotein polypeptides B and B1

GATA4 GATA binding protein 4 DCM, AF

DPP6 dipeptidyl-peptidase 6

F7 coagulation factor VII (serum prothrombin conversion accelerator)

CNOT1 CCR4-NOT transcription complex, subunit 1

KCNJ11 potassium inwardly-rectifying channel, subfamily J, member 11

LITAF lipopolysaccharide-induced TNF factor

ABCC8 ATP-binding cassette, sub-family C (CFTR/MRP), member 8 FH

CCT8 chaperonin containing TCP1, subunit 8 (theta)

HNRNPK heterogeneous nuclear ribonucleoprotein K

TCP1 t-complex 1

CNOT4 CCR4-NOT transcription complex, subunit 4

UBE4A ubiquitination factor E4A

385 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

Gene Gene Name ICC Diseases

SRL

TRIM54 tripartite motif containing 54 HCM

ZBTB17 zinc finger and BTB domain containing 17

UBE2D1 ubiquitin-conjugating enzyme E2D 1

IL18 interleukin 18 (interferon-gamma-inducing factor)

PFDN5 prefoldin subunit 5

PFDN2 prefoldin subunit 2

CAPN1 1, (mu/I) large subunit

HSP90AA1 heat shock protein 90kDa alpha (cytosolic), class A member 1

ATP2A2 ATPase, Ca++ transporting, cardiac muscle, slow twitch 2

RANGRF RAN guanine nucleotide release factor BrS

NOS1AP nitric oxide synthase 1 (neuronal) adaptor protein

SNRPA small nuclear ribonucleoprotein polypeptide A

FBXO32 F-box protein 32

LIPC lipase, hepatic FH, FHL

SELP selectin P (granule membrane protein 140kDa, antigen CD62)

POLR2F polymerase (RNA) II (DNA directed) polypeptide F

GINS3 GINS complex subunit 3 (Psf3 homolog)

UBE2D2 ubiquitin-conjugating enzyme E2D 2

ATP1B1 ATPase, Na+/K+ transporting, beta 1 polypeptide

CALM1 calmodulin 1 (phosphorylase kinase, delta) LQTS, CPVT

CCT7 chaperonin containing TCP1, subunit 7 (eta)

PFDN4 prefoldin subunit 4

GAA glucosidase, alpha; acid HCM

DNAJC1 DnaJ (Hsp40) homolog, subfamily C, member 1

UBE2D4 ubiquitin-conjugating enzyme E2D 4 (putative)

RBM46 RNA binding motif protein 46

ENDOG endonuclease G

HNRNPU heterogeneous nuclear ribonucleoprotein U (scaffold attachment factor A)

RXRB retinoid X receptor, beta

386 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

8.2 Standard Operating Protocol for Date of Clinical Diagnosis of DCM

Date of clinical diagnosis of DCM

In chronological order, the first date the following findings are noted is the date of diagnosis. In order of preference: 1. Earliest date of hospital-based specialist (cardiologist) diagnosis regardless of assessment modality a. Examples: Cardiology clinic letter, discharge summary (any team), MDT report b. Does not include: GP referrals, non-specialist letters c. Diagnosis terms in order of preference: i. Dilated cardiomyopathy ii. Heart failure/Congestive cardiac failure – only if accompanied by imaging evidence of dilated and impaired LV

2. Earliest date of isolated investigation confirming diagnosis (in order of preference) a. Cardiac MRI specifically referring to diagnosis of dilated cardiomyopathy b. Coronary angiogram specifically referring to diagnosis of dilated cardiomyopathy c. Cardiac MRI with gadolinium contrast showing dilated and impaired LV in absence of infarct pattern d. Coronary angiogram demonstrating normal coronary arteries with recent (<3 months) echocardiogram showing dilated and impaired LV e. Echocardiogram showing dilated and impaired LV

In the absence of any of the above: 3. GP summary referring to date of diagnosis of (in order of preference) a. Dilated cardiomyopathy b. Heart failure/Congestive cardiac failure c. Cardiomyopathy (e.g. Peri-partum cardiomyopathy) d. Excludes: Hypertrophic cardiomyopathy, ischaemic cardiomyopathy

387 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

8.3 CMR safety checklist

Figure 8-1: CMR safety checklist

388 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

8.4 Endpoint definitions

8.4.1.1 Endpoint definitions

Endpoints were defined according to the 2014 American College of Cardiology/American Heart Association definitions for cardiovascular endpoints in clinical trials(401).

8.4.1.1.1 Cardiovascular death

Cardiovascular death was defined as death due to sudden cardiac death, heart failure, acute myocardial infarction, cerebrovascular accident, cardiovascular haemorrhage, cardiovascular procedures, or other cardiovascular causes, that is death not included in the previous categories but with a specific, known cause such as pulmonary embolus (Table 8-1).

Table 8-1: Definition of cardiovascular death. Taken from the 2014 American College of Cardiology and American Heart Association definition of cardiovascular endpoints in clinical trials document(401). CV=cardiovascular, MI= myocardial infarction, HF= heart failure, ICD=implantable cardiac defibrillator.

Acute Death by any cardiovascular mechanism (arrhythmia, sudden death, heart myocardial failure, stroke, pulmonary embolus, peripheral arterial disease) within 30 d after infarction an acute MI, related to the immediate consequences of the MI, such as (MI) progressive HF or recalcitrant arrhythmia. There may be assessable (attributable) mechanisms of cardiovascular death during this time period, but for simplicity, if the cardiovascular death occurs within 30 d of an acute MI, it will be considered a death due to MI. Note: Acute MI should be verified to the extent possible by the diagnostic criteria outlined for acute MI or by autopsy findings showing recent MI or recent coronary thrombosis. Death resulting from a procedure to treat an MI (PCI or CABG), or to treat a complication resulting from MI, should also be considered death due to acute MI. Death resulting from an elective coronary procedure to treat myocardial ischemia (ie, chronic stable angina) or death due to an MI that occurs as a direct consequence of a cardiovascular investigation/procedure/operation should be considered as a death due to a cardiovascular procedure. Sudden Death that occurs unexpectedly and not within 30 d of an acute MI. cardiac Note: Sudden cardiac death includes the following scenarios: death • Death witnessed and occurring without new or worsening symptoms • Death witnessed within 60 min of the onset of new or worsening cardiac symptoms unless the symptoms suggest acute MI • Death witnessed and attributed to an identified arrhythmia (eg, captured on an electrocardiographic recording, witnessed on a monitor, or unwitnessed but found on ICD review) • Death after unsuccessful resuscitation from cardiac arrest (eg, ICD unresponsive sudden cardiac death, pulseless electrical activity arrest) • Death after successful resuscitation from cardiac arrest and without identification of a specific cardiac or noncardiac etiology • Unwitnessed death in a subject seen alive and clinically stable ≤24 h before being found dead without any evidence supporting a specific noncardiovascular cause of death (information about the patient’s clinical status preceding death

389 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

should be provided if available) Unless additional information suggests an alternate specific cause of death (eg, Death due to Other Cardiovascular Causes), if a patient is seen alive ≤24 h before being found dead, sudden cardiac death should be recorded. For patients who were not observed alive within 24 h of death, undetermined cause of death should be recorded (eg, a subject found dead in bed but who had not been seen by family members for >24 h). Heart failure Death associated with clinically worsening symptoms and/or signs of HF, regardless of HF etiology Note: Deaths due to HF can have various etiologies, including single or recurrent MIs, ischemic or nonischemic cardiomyopathy, hypertension, or valvular disease.

Stroke Death after a stroke that is either a direct consequence of the stroke or a complication of the stroke. Note: Acute stroke should be verified to the extent possible by the diagnostic criteria outlined for stroke.

CV Death caused by the immediate complication(s) of a Cardiovascular procedure procedure

CV Death related to hemorrhage such as a nonstroke intracranial hemorrhage, (eg, hemorrhage subdural hematoma) nonprocedural or nontraumatic vascular rupture (eg, aortic aneurysm), or hemorrhage causing cardiac tamponade

CV other Cardiovascular death not included in the above categories but with specific, known cause (eg, pulmonary embolus)

8.4.1.1.2 Heart Failure composite

The heart failure composite consisted of either a heart transplant, left ventricular assist-device implantation or an unplanned heart failure hospitalisation.

An unplanned heart failure hospitalisation was defined as ‘an event in which the patient is admitted to the hospital with a primary diagnosis of heart failure, the length of stay is at least 24 h (or extends over a calendar date), the patient exhibits new or worsening symptoms of heart failure on presentation, has objective evidence of new or worsening heart failure, and receives initiation or intensification of treatment specifically for heart failure’(401). Notably, changes to oral diuretic therapy did not qualify as initiation or intensification of treatment(401). New or worsening symptoms due to heart failure consisted of at least one of dyspnea, decreased exercise tolerance, fatigue, worsened end organ perfusion or volume overload(401).

Objective evidence of heart failure consisted of at least 2 physical examination findings or at least 1 physical examination finding and at least 1 laboratory criterion of new or worsening heart failure on presentation(401).

Physical examination findings included new or worsening peripheral oedema, ascites, pulmonary crepitations, increased jugular venous pressure, S3 gallop, or clinically significant weight gain judged to be secondary to fluid retention(401). Laboratory criteria included new

390 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

or worsening BNP/NT-proBNP, radiological evidence of pulmonary congestion, or other invasive or non invasive diagnostic evidence of heart failure(401).

8.4.1.1.3 Arrhythmia composite

All ventricular arrhythmias were defined in line with the American College of Cardiology, American Heart Association and European Society of Cardiology guidelines(416). The arrhythmia endpoint was a composite of sustained ventricular tachycardia, ventricular fibrillation, aborted sudden cardiac death and appropriate implantable cardioverter- defibrillator (ICD) discharge.

Sustained ventricular tachycardia was defined as repetitive ventricular beats in a row lasting over 30 seconds in duration at a rate greater than 100 beats per minute (cycle length less than 600 ms), and/or requiring termination due to hemodynamic compromise in less than 30 seconds(416).

Ventricular fibrillation was defined as a rapid, usually more than 300 bpm (cycle length 180 ms or less), grossly irregular ventricular rhythm with marked variability in QRS cycle length, morphology, and amplitude(416).

Aborted sudden cardiac death was diagnosed if patients received an appropriate ICD shock for ventricular arrhythmia, or had a nonfatal episode of ventricular fibrillation or spontaneous sustained ventricular tachycardia (>30 seconds in duration) causing hemodynamic compromise and requiring cardioversion(13,416). Appropriate ICD discharge was limited to shock therapy for ventricular arrhythmias (anti-tachycardia pacing and inappropriate shocks for atrial or other arrhythmias were excluded) and all discharges were adjudicated by an Electrophysiology Cardiology Consultant.

391 Template by Friedman & Morgan 2014 Word Chapter 8: Appendices

8.5 Permissions

392 Template by Friedman & Morgan 2014 Word