1 GENETIC DETERMINANTS AND GENOTYPE- PHENOTYPE ASSOCIATIONS IN HYPERTROPHIC CARDIOMYOPATHY - CONTRIBUTION OF A HIGH-THROUGHPUT SEQUENCING APPROACH LUÍS MIGUEL DA ROCHA LOPES UNIVERSITY COLLEGE LONDON, UCL DOCTOR OF PHILOSOPHY (PH.D.) 2015
385
Embed
GENETIC DETERMINANTS AND GENOTYPEPHENOTYPE ASSOCIATIONS IN HYPERTROPHIC CARDIOMYOPATHY - CONTRIBUTION OF A HIGH-THROUGHPUT SEQUENCING APPROACH
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
CARDIOMYOPATHY - CONTRIBUTION OF A HIGH-THROUGHPUT SEQUENCING APPROACH UNIVERSITY COLLEGE LONDON, UCL DOCTOR OF PHILOSOPHY (PH.D.) 2015 2 DECLARATION I, Luís Miguel da Rocha Lopes, confirm that the work presented in this thesis is my own. Where information has been derived from other sources, I confirm that this has been indicated in the thesis. This thesis was evaluated using the online tool “Turnitin”, accessed within UCL Moodle. The similarity index was 19%. According to the originality report, this percentage is due to already published work by the PhD candidate as the first author (Appendix A). Some parts of the methodology used to obtain the data analysed in this work have been developed in collaboration with the following investigators, and this is indicated where appropriate throughout: - Doctor Mike Hubank, Institute of Child Health, UCL: “Targeted gene enrichment and high-throughput sequencing of 41 cardiovascular genes” – “Library design and capture and sequencing protocol optimization” (Methods, Chapter 2); “Sequence data analysis of the non-coding regions” – “Expression study in patient samples vs controls” (Methods, Chapter 7); - Doctor Vincent Plagnol, UCL Genetics Institute, UCL: “Sequence data analysis for targeted exonic regions” – “Bioinformatic analysis pipeline” (Methods, Chapter 3) and “Screening for copy number variation” – “Bioinformatic analysis for the detection of copy-number variation – ExomeDepth” (Methods, Chapter 6); - Professor Mathias Gautel, King’s College London: “Titin variants” – “Manual annotation and prediction of pathogenicity “ (Methods, Chapter 3); - Doctor Chiara Bachelli, GOSGene, Institute of Child Health, UCL: “Whole-exome sequencing and data analysis of sarcomere-negative families” (Methods, Chapter 8); - Doctor Andrew Martin, Structural and Molecular Biology, UCL: “In silico analysis of MYH7 missense variants – pathogenicity and phenotype prediction based in the structural impact of the mutation” –“Machine learning analysis to predict phenotype” (Methods, Chapter 9). 3 ACKNOWLEDGMENTS I was supported by a grant from the Gulbenkian Doctoral Programme for Advanced Medical Education, sponsored by Fundação Calouste Gulbenkian, Ministério da Saúde and Fundação para a Ciência e Tecnologia, Portugal. I am deeply in depth to the Gulbenkian Foundation for this grant, which allowed me to enrol and conduct my research at University College London and for the initial 6 months of graduate courses, which influenced and inspired many ideas for the following 4 years. A special word of gratitude to Professors Leonor Parreira and João Ferreira, for their teaching, wise support and encouragement. I am profoundly glad for the opportunity to thank my primary supervisor, Professor Perry Elliott. For his time, availability, encouragement, wisdom, patience and continued mentorship and example. Also for allowing me the ambition of participating in such a large and challenging project, which transformed this period of my life in a continuous learning experience. I also owe my supervisor a dept of gratitude for what was an intense period of clinical learning in the field of inherited heart disease. I am also very grateful to my secondary supervisor, Professor William McKenna, for his encouragement, example, wisdom and intellectual contribution to my research and writing. A special word of deep gratitude to Doctor Petros Syrris for his availability, teaching, friendship, support, wise advice and continuous encouragement along all this period. I also thank all the collaborators of this project from outside the Institute of Cardiovascular Science, for their generous contribution and time and from whom I learned immensely in different areas of the Life Sciences. In particular, Doctor Vincent Plagnol, Doctor Mike Hubank, Doctor Andrew Martin and Professor Mathias Gautel. I thank all my friends, fellows, graduate students, consultants, nurses and genetic counsellors at The Heart Hospital, for their help and companionship. Without the work of all present and past colleagues, my research project would not have been possible. I am also deeply grateful to all the patients and families who participated in this study and from whom I have learned so much. I would like to dedicate this thesis to my parents, for their lifelong support and encouragement. 4 ABSTRACT The application of hypertrophic cardiomyopathy (HCM) genetics in clinical practice has been limited by an incomplete knowledge of the genetic background and a poor understanding of genotype-phenotype relationships. The aims of this study were to study genotype-phenotype relationships in HCM, expand the knowledge on the genetic architecture of the disease, explore genetic modifiers of the phenotype and develop a methodology for interpretation of variants detected by high-throughput sequencing platforms. Methods The study population consisted in consecutive and unrelated HCM patients. In order to analyze coding, intronic and regulatory regions of 41 cardiovascular genes, solution-based sequence capture was followed by massive parallel resequencing. Single-nucleotide variants, small insertion/deletions and copy number variants (CNVs) were called. For the analysis of variants in the coding region, rare, non-synonymous, loss-of-function and splice-site variants were defined as candidates. These variants were tested for associations with clinical phenotype and survival. For the analysis of non-coding variation, variants located in known transcription factor (TF) binding sites and 3’UTR miRNA targets were identified. The performance of an insilico pathogenicity prediction strategy incorporating structural features was explored for MYH7 variants. Selected sarcomere-negative families were studied by whole-exome sequencing (WES). Results Eight-hundred-and-seventy-four patients (49.6±15.4 years, 67.8% males) were studied; likely disease-causing sarcomere protein (SP) gene variants were detected in 383 (43.8%). Patients with SP variants were characterized by younger age and higher prevalence of family history of HCM, family history of sudden cardiac death, asymmetric septal hypertrophy, higher maximal wall thickness (P-values<0.0005) and increased cardiovascular death (P-value=0.012). Similar associations were observed for individual SP genes. Patients with ANK2 variants had greater maximum wall thickness (P-value=0.0005). Associations at a lower level of significance were demonstrated with variation in other non-SP genes. Four CNVs were detected in MYBPC3, PDLIM3, TNNT2 and LMNA. Fourteen percent carried non-coding variants mapping to TF 5 binding sites. The pathogenicity prediction for MYH7 missense variants had an accuracy of 0.93; the phenotype predictor had an accuracy of 0.79 and novel genotype-phenotype associations were suggested. Three families were studied using WES; in two of these families the main candidate variants were in OBSCN and TTN. Conclusions Patients with SP variants differed from those without with respect to age, family history, ventricular morphology and prognosis. Novel associations were demonstrated between individual sarcomeric genes and several phenotype traits and for the first time, associations between rare variants in non-SP genes and phenotype were described. CNVs and non-coding variation in SP genes can additionally contribute to the genetic architecture of HCM. Pathogenicity prediction incorporating structural features revealed additional genotype- phenotype associations. Whole-exome sequencing suggested new causal genes. 6 TABLE OF CONTENTS DECLARATION ............................................................................................................................... 2 ACKNOWLEDGMENTS ..................................................................................................................... 3 ABSTRACT .................................................................................................................................... 4 LIST OF TABLES ........................................................................................................................... 10 LIST OF FIGURES .......................................................................................................................... 12 LIST OF ABBREVIATIONS ............................................................................................................... 14 SECTION I – INTRODUCTION ...................................... 17 1. DEFINITION, PHENOTYPE AND EPIDEMIOLOGY OF HYPERTROPHIC CARDIOMYOPATHY ....................... 18 2. GENETIC BACKGROUND AND MECHANISMS OF DISEASE .............................................................. 20 2.1. Normal sarcomere structure ....................................................................................... 20 2.2. Mechanism of contraction – the cross-bridge cycle .................................................... 22 2.3. Contractile protein gene mutations ............................................................................ 24 2.4. Role of the non-contractile sarcomere proteins in disease ......................................... 26 2.4.1. Titin .................................................................................................................................................... 26 2.4.2. Z-disc and associated proteins ........................................................................................................... 27 2.4.3. M-band and associated proteins ....................................................................................................... 29 2.5. Downstream mechanisms of disease and possible drug targets ................................ 32 2.5.1. Cross-bridge kinetics .......................................................................................................................... 32 2.5.2. Calcium sensitivity and cycling ........................................................................................................... 33 2.5.3. Signaling pathways and protein degradation pathways .................................................................... 34 2.5.4. Cardiomyocyte-fibroblast cross-talk: fibrosis..................................................................................... 35 2.5.5. Cardiomyocyte energetics ................................................................................................................. 36 3. COMPLEXITY AND CHALLENGES IN HYPERTROPHIC CARDIOMYOPATHY GENETICS.............................. 38 3.1. Genetic testing strategies, high-throughput sequencing and interpretation of rare variation ............................................................................................................................. 38 3.2. Genotype-phenotype associations .............................................................................. 40 3.3. Structural variation ..................................................................................................... 42 3.4. Non-coding variation ................................................................................................... 42 3.5. In silico prediction of the pathogenicity of a coding variant ....................................... 43 4. SYSTEMATIC REVIEW OF THE LITERATURE AND META-ANALYSIS OF GENOTYPE-PHENOTYPE ASSOCIATIONS.. .......................................................................................................................... 45 4.1. Methods ...................................................................................................................... 45 4.1.1. Study selection and electronic search methods ................................................................................ 45 4.1.2. Statistical analysis .............................................................................................................................. 46 4.2. Results ......................................................................................................................... 47 4.2.1. Demographic characteristics and family history ................................................................................ 53 4.2.2. Morphology and function .................................................................................................................. 59 4.2.3. Risk factors for sudden cardiac death ................................................................................................ 63 4.2.4. Interventions and prognosis .............................................................................................................. 65 4.3. Summary of results ...................................................................................................... 67 5. AIMS ................................................................................................................................. 69 6. HYPOTHESES ....................................................................................................................... 69 SECTION II – METHODS ........................................... 70 1. STUDY COHORT AND ETHICAL APPROVAL .................................................................................. 71 2. TARGETED GENE ENRICHMENT AND HIGH-THROUGHPUT SEQUENCING OF 41 CARDIOVASCULAR GENES….. .................................................................................................................................. 72 2.1. Phase I – Library design and capture and sequencing protocol optimization ............. 74 2.2. Phase II – Sequencing of the study cohort ................................................................... 75 7 2.3. Sequencing and analysis of UK10K control samples ................................................ 76 3. SEQUENCE DATA ANALYSIS FOR TARGETED EXONIC REGIONS ........................................................ 78 3.1. Bioinformatic analysis pipeline .................................................................................... 78 3.2. Manual annotation of the variants and in silico prediction of pathogenicity ............. 79 3.3. Titin variants - manual annotation and prediction of pathogenicity .......................... 80 3.4. Analysis of UK10K control samples and case-control comparison of candidate nsSNPs ............................................................................................................................................ 81 3.5. Case control comparison of candidate nsSNPs with UCL-exomes cohort ................... 82 4. CLINICAL VALIDATION AND ASSESSMENT OF THE ACCURACY OF THE HIGH-THROUGHPUT SEQUENCING PLATFORM AND ANALYSIS PIPELINE ................................................................................................. 83 4.1. Step 1 –Systematic validation of accuracy for plates 1-3 (first 223 patients) ............. 83 4.2. Step 2 – Application - validation of clinically actionable variants (all patients) ......... 84 5. PHENOTYPE DATA COLLECTION AND GENOTYPE-PHENOTYPE ANALYSIS .......................................... 85 5.1. Demographic data and symptoms .............................................................................. 85 5.2. Electrocardiography .................................................................................................... 85 5.3. Cardio-pulmonary exercise testing .............................................................................. 86 5.4. Ambulatory ECG monitoring ....................................................................................... 86 5.5. Echocardiography ........................................................................................................ 86 5.6. Definition of the clinical risk factors for sudden cardiac death ................................... 87 5.7. Cardiac magnetic resonance protocol ......................................................................... 87 5.8. Statistical analysis ....................................................................................................... 88 5.8.1. Descriptive statistics and comparison of means and proportions between groups .......................... 88 5.8.2. Multiple comparison correction strategy .......................................................................................... 89 5.8.3. Logistic regression analysis and construction of a model and score to predict the presence of a sarcomere protein gene mutation ............................................................................................................... 89 5.8.4. Survival analysis ................................................................................................................................. 90 6. SCREENING FOR COPY NUMBER VARIATION .............................................................................. 91 6.1. Study population ......................................................................................................... 91 6.2. Bioinformatic analysis for the detection of copy-number variation – ExomeDepth ... 91 6.3. Array comparative genomic hybridization .................................................................. 92 7. ANALYSIS OF THE NON-CODING REGIONS ................................................................................. 93 7.1. Study cohort ............................................................................................................. 93 7.2. Analysis of the non-coding variants ......................................................................... 93 7.3. Expression study in patient samples versus controls ............................................... 94 8. WHOLE-EXOME SEQUENCING AND DATA ANALYSIS OF SARCOMERE-NEGATIVE FAMILIES................... 96 8.1. Samples ....................................................................................................................... 96 8.2. GOSgene application and collaboration ...................................................................... 96 8.3. DNA quantification and quality ................................................................................... 97 8.4. Exome capture and sequencing ................................................................................... 97 8.5. Exome analysis ............................................................................................................ 97 8.6. Filtering and prioritization of variants......................................................................... 98 9. IN SILICO ANALYSIS AND STRUCTURAL IMPACT PREDICTION OF MISSENSE VARIANTS IN MYH7 ........ 101 9.1. Dataset of MYH7 variants ......................................................................................... 101 9.2. Prediction of in silico pathogenicity........................................................................... 102 9.3. Manual analysis of associations between structural consequences and phenotype 102 9.4. Machine learning analysis to predict pathogenicity and phenotype ........................ 103 SECTION III – RESULTS ............................................ 105 1. STUDY COHORT CHARACTERIZATION ...................................................................................... 106 2. TARGETED GENE ENRICHMENT AND HIGH-THROUGHPUT SEQUENCING – COVERAGE AND READ DEPTH DATA. ...................................................................................................................................... 109 2.1. Analysis per plate ...................................................................................................... 109 2.2. Analysis per gene ....................................................................................................... 112 3. CLINICAL VALIDATION AND ASSESSMENT OF THE ACCURACY OF THE HIGH-THROUGHPUT SEQUENCING PLATFORM AND ANALYSIS PIPELINE ............................................................................................... 114 3.1. Step 1 – Systematic validation of accuracy for plates 1-3 (first 223 patients) .......... 114 3.2. Step 2 – Application - validation of clinically actionable variants (all patients) ....... 114 4. CODING SEQUENCE DATA ANALYSIS ....................................................................................... 115 4.1. Sarcomere protein gene variants .............................................................................. 115 4.2. Other sarcomere and associated genes .................................................................... 122 4.3. First case – control analysis: comparison of nsSNPs between the first 223 HCM cases and UK10K controls .......................................................................................................... 124 4.4. Second case – control analysis: comparison of all types of variants between all the 874 HCM cases and UCL-exome controls ......................................................................... 127 4.5. Recurrent variants in HCM cases ............................................................................... 129 4.6. Titin variants .............................................................................................................. 130 4.7. Non-sarcomere protein gene variants ....................................................................... 136 5. PHENOTYPE DATA AND GENOTYPE-PHENOTYPE ANALYSIS .......................................................... 140 5.1. Effect of mutations in sarcomere genes .................................................................... 140 5.2. Patients with multiple sarcomere protein gene variants .......................................... 141 5.3. Associations with rare variants in desmosomal and ion channel genes ................... 141 5.4. Comparisons between sarcomeric genes .................................................................. 141 5.5. Genotype-phenotype associations for variants predicted in silico to be pathogenic 142 5.6. Comparisons within sarcomere positive individuals only .......................................... 142 5.7. Logistic regression analysis and construction of a model and score to predict the presence of a sarcomere gene mutation .......................................................................... 152 5.8. Titin genotype-phenotype associations ..................................................................... 153 5.8.1. Comparison of means and proportions for the different phenotypic traits .................................... 153 5.8.2. Discriminant analysis and logistic regression analysis ..................................................................... 154 6. CARDIAC MAGNETIC RESONANCE IMAGING PHENOTYPE DATA .................................................... 157 7. SCREENING OF COPY NUMBER VARIATION .............................................................................. 160 8. NON-CODING REGION ANALYSIS ........................................................................................... 172 8.1. Variation in transcription factor binding sites ........................................................... 172 8.2. Normalizing the number of distinct non-coding variants for the size of each gene . 175 8.3. Evolutionary conservation analysis of the genomic coordinates .............................. 175 8.4. Transcription factors previously associated with cardiomyocyte hypertrophy and cardiomyopathy signaling pathways ............................................................................... 175 8.5. 3’UTR variants in miRNA target sites ........................................................................ 177 8.6. Expression comparison between cases and controls ................................................ 178 9. WHOLE-EXOME SEQUENCING FOR SARCOMERE-NEGATIVE FAMILIES ........................................... 180 9.1. Quality assessment of the sequencing data ........................................................... 180 9.2. Candidate variants ................................................................................................. 180 10. IN SILICO ANALYSIS OF MYH7 MISSENSE VARIANTS – PATHOGENICITY AND PHENOTYPE PREDICTION BASED IN THE STRUCTURAL IMPACT OF THE MUTATION .................................................................... 183 10.1. Manual analysis ...................................................................................................... 183 10.2. Machine learning analysis to predict pathogenicity ............................................... 184 10.3. Machine learning analysis to predict phenotype (HCM vs DCM) ............................ 185 SECTION IV – DISCUSSION ....................................... 187 9 1.2.1. Sarcomere and related genes .......................................................................................................... 189 1.2.2. Non-sarcomere genes ...................................................................................................................... 192 1.3. Determining pathogenicity of sequence variants...................................................... 193 2. NOVEL GENOTYPE-PHENOTYPE ASSOCIATIONS IN HYPERTROPHIC CARDIOMYOPATHY REVEALED BY HIGH- THROUGHPUT SEQUENCING ........................................................................................................ 197 2.1. Influence of sarcomeric variation on phenotype ....................................................... 197 2.2. Cardiac magnetic resonance phenotyping suggested additional genotype-phenotype associations ...................................................................................................................... 199 2.3. Effects of titin on the phenotype ............................................................................... 200 2.4. Modifier effect of non-sarcomere variants ............................................................... 201 2.5. Clinical implications ................................................................................................... 202 3. COPY NUMBER VARIATION IN HYPERTROPHIC CARDIOMYOPATHY ............................................... 203 4. NON-CODING SEQUENCE DATA ANALYSIS ............................................................................... 206 4.1. Variants mapping to transcription factor binding sites ............................................ 206 4.2. Expression comparison between cases and controls ................................................ 208 5. WHOLE-EXOME SEQUENCING .............................................................................................. 210 6. IN SILICO ANALYSIS OF MYH7 MISSENSE VARIANTS – PATHOGENICITY AND PHENOTYPE PREDICTION BASED IN THE STRUCTURAL IMPACT OF THE MUTATION .................................................................... 213 7. CONCLUSIONS ................................................................................................................... 215 REFERENCES ......................................................... 216 APPENDICES ......................................................... 237 APPENDIX A – PUBLICATIONS ARISING FROM THIS WORK ................................................................. 238 APPENDIX B – ETHICAL APPROVAL AND CONSENT FORM .................................................................. 240 APPENDIX C – ADDITIONAL STATISTICAL METHODS FOR METHODS - CHAPTER 3.4 ............................... 253 APPENDIX D – LIST OF CANDIDATE VARIANTS – EXCLUDING TITIN ...................................................... 258 APPENDIX E – LIST OF CANDIDATE TITIN VARIANTS ......................................................................... 315 APPENDIX F – LISTS OF CANDIDATE VARIANTS FROM WHOLE-EXOME SEQUENCING .............................. 367 APPENDIX G – LIST OF MYH7 VARIANTS USED FOR PATHOGENICITY AND PHENOTYPE PREDICTION BASED IN THE STRUCTURAL IMPACT OF THE MUTATION ................................................................................. 375 10 LIST OF TABLES Table 1. Summary of the sarcomeric and associated proteins, encoding genes, interactions, associated disease phenotypes and respective prevalence 31 Table 2. Family centred studies reporting on penetrance, severity and prognosis 49 Table 3. Studies selected for pooled analyses. Frequency of mutations in each sarcomere gene and sequencing methodology for each study 52 Table 4. Name of the targeted genes, Ensembl accession number, chromosomal position and size 73 Table 5. Demographic and clinical characteristics of the study cohort 106 Table 6. Pooled analysis of read depth per gene, across all plates and subdivided in plates 112 Table 7. Prevalence of rare variants in the eight main sarcomere genes 116 Table 8. Number of distinct rare variants in sarcomeric, Z-disc and calcium-handling genes 123 Table 9. Level of evidence for the pathogenicity of the distinct variants 124 Table 10. Rare nsSNPs frequency comparison between my sequencing results and a set of 1,287 UK controls with exome sequence data generated by the UK10K project for 19 HCM/DCM associated genes 126 Table 11. Case-control comparison for the frequency of rare (MAF<0.2%) variants between the 874 cases vs UCL-exome samples as controls for sarcomere and associated genes 128 Table 12. Candidate variants present in the first 223 HCM cases for which the single nsSNP case- control p-value between HCM cases and UK10K controls was < 0.05 130 Table 13. Candidate variants present in 874 HCM cases for which the single case-control P-value between HCM cases and UCL-exome controls was < 0.05 130 Table 14. Prioritized TTN variants and predicted effect 133 Table 15. Comparison between the frequency of individual TTN variants in this cohort and the 1000 genomes population 134 Table 16. Number of distinct rare variants in genes associated with arrhythmogenic cardiomyopathy and ion channel disease 137 Table 17. Rare nsSNPs frequency comparison between my sequencing results and a set of 1,287 UK controls with exome sequence data generated by the UK10K project for genes associated with arrhythmogenic right ventricular cardiomyopathy and ion channel disease 138 Table 18. Case-control comparison for the frequency of rare (MAF<0.2%) variants between the 874 HCM cases vs UCL-exome samples as controls for ARVC associated and ion-channel genes 139 Table 19. Genotype-phenotype associations for individual sarcomeric protein genes and non- sarcomere protein genes, meeting the predefined statistical thresholds for multiple testing 144 Table 20. Additional genotype-phenotype associations for individual sarcomere protein and related genes, with P-values <0.05, but not meeting the predefined threshold for significance 146 Table 21. Genotype-phenotype associations for non-sarcomeric protein genes not meeting the predefined statistical thresholds for multiple testing 151 Table 22. Multivariate logistic regression model for the presence of a rare variant in a SP gene 152 Table 23. Multivariate logistic regression analysis for the presence of a potentially truncating titin variant 155 Table 24. Multivariate logistic regression analysis for the presence of an enriched titin variant 156 Table 25. Characterization of the subset of the total cohort studied with CMR 157 Table 26. Cardiac magnetic resonance parameters 158 11 Table 27. Comparison between sarcomere gene mutation-positive and negative patients for the analysed CMR parameters 159 Table 28. Demographic, clinical and genetic characteristics of the patients harbouring CNVs 161 Table 29. List of non-coding variants that map onto transcription factors binding sites 173 Table 30. Rare non-coding variants that map onto predicted miRNA target regions 178 Table 31. Raw and normalized values of mRNA expression of the two cases and the controls for the eight main sarcomere genes 179 Table 32. Depth of coverage statistics for whole-exome sequencing 180 12 Figure 1. Schematic showing the different main cardiomyopathy phenotypes 18 Figure 2. Schematic representation of the main eight sarcomere proteins and their interactions in the sarcomere 23 Figure 3. Schematic representation of the discrete topographic zones in each sarcomere 24 Figure 4. Schematic representation of the sarcomere, focusing on the Z-disc and the M-band proteins and its interactions 30 Figure 5. Mechanisms of disease in hypertrophic cardiomyopathy and possible drug targets 37 Figure 6. Flow chart of study selection process for the meta-analysis 47 Figure 7. Pooled analysis for the proportion of individuals that are sarcomere gene mutation-positive 53 Figure 8. Forest plot from random effect meta-analysis showing that the presence of a sarcomere gene mutation was associated with a younger age at presentation 54 Figure 9. Forest plot from random effect meta-analysis showing no statistically significant difference regarding the age of presentation between MYBPC3 and MYH7 55 Figure 10. Forest plot from random effect meta-analysis showing the absence of a statistically significant difference between the proportion of males in sarcomere-positive compared with sarcomere-negative patients 56 Figure 11. Forest plot from random effect meta-analysis showing a significantly…