Details

Individual Participant Data Meta-Analysis


Individual Participant Data Meta-Analysis

A Handbook for Healthcare Research
Statistics in Practice 1. Aufl.

von: Richard D. Riley, Jayne F. Tierney, Lesley A. Stewart

67,99 €

Verlag: Wiley
Format: EPUB
Veröffentl.: 24.05.2021
ISBN/EAN: 9781119333753
Sprache: englisch
Anzahl Seiten: 560

DRM-geschütztes eBook, Sie benötigen z.B. Adobe Digital Editions und eine Adobe ID zum Lesen.

Beschreibungen

<p><i>Individual Participant Data Meta-Analysis: A Handbook for Healthcare Research</i> provides a comprehensive introduction to the fundamental principles and methods that healthcare researchers need when considering, conducting or using individual participant data (IPD) meta-analysis projects. Written and edited by researchers with substantial experience in the field, the book details key concepts and practical guidance for each stage of an IPD meta-analysis project, alongside illustrated examples and summary learning points.</p> <p>Split into five parts, the book chapters take the reader through the journey from initiating and planning IPD projects to obtaining, checking, and meta-analysing IPD, and appraising and reporting findings. The book initially focuses on the synthesis of IPD from randomised trials to evaluate treatment effects, including the evaluation of participant-level effect modifiers (treatment-covariate interactions). Detailed extension is then made to specialist topics such as diagnostic test accuracy, prognostic factors, risk prediction models, and advanced statistical topics such as multivariate and network meta-analysis, power calculations, and missing data.</p> <p>Intended for a broad audience, the book will enable the reader to:</p> <ul> <li>Understand the advantages of the IPD approach and decide when it is needed over a conventional systematic review</li> <li>Recognise the scope, resources and challenges of IPD meta-analysis projects</li> <li>Appreciate the importance of a multi-disciplinary project team and close collaboration with the original study investigators</li> <li>Understand how to obtain, check, manage and harmonise IPD from multiple studies</li> <li>Examine risk of bias (quality) of IPD and minimise potential biases throughout the project</li> <li>Understand fundamental statistical methods for IPD meta-analysis, including two-stage and one-stage approaches (and their differences), and statistical software to implement them</li> <li>Clearly report and disseminate IPD meta-analyses to inform policy, practice and future research</li> <li>Critically appraise existing IPD meta-analysis projects</li> <li>Address specialist topics such as effect modification, multiple correlated outcomes, multiple treatment comparisons, non-linear relationships, test accuracy at multiple thresholds, multiple imputation, and developing and validating clinical prediction models</li> </ul> Detailed examples and case studies are provided throughout.
<p>Acknowledgements xxiii</p> <p><b>1 Individual Participant Data Meta-analysis for Healthcare Research </b><b>1<br /></b><i>Richard D. Riley, Lesley A. Stewart, and Jayne F. Tierney</i></p> <p>1.1 Introduction 1</p> <p>1.2 What Is IPD and How Does It Differ from Aggregate Data? 1</p> <p>1.3 IPD Meta-analysis: A New Era for Evidence Synthesis 2</p> <p>1.4 Scope of This Book and Intended Audience 2</p> <p><b>Part I Rationale, Planning, and Conduct </b><b>7</b></p> <p><b>2 Rationale for Embarking on an IPD Meta-analysis Project </b><b>9<br /></b><i>Jayne F. Tierney, Richard D. Riley, Catrin Tudur Smith, Mike Clarke, and Lesley A. Stewart</i></p> <p>2.1 Introduction 9</p> <p>2.2 How Does the Research Process Differ for IPD and Aggregate Data Meta-analysis Projects? 10</p> <p>2.2.1 The Research Aims 10</p> <p>2.2.2 The process and methods 10</p> <p>2.3 What Are the Potential Advantages of an IPD Meta-analysis Project? 11</p> <p>2.4 What Are the Potential Challenges of an IPD Meta-Analysis Project? 14</p> <p>2.5 Empirical Evidence of Differences between Results of IPD and Aggregate Data Metaanalysis Projects 14</p> <p>2.6 Guidance for Deciding When IPD Meta-analysis Projects Are Needed to Evaluate Treatment Effects from Randomised Trials 15</p> <p>2.6.1 Are IPD Needed to Tackle the Research Question? 15</p> <p>2.6.2 Are IPD Needed to Improve the Completeness and Uniformity of Outcomes and Participant-level Covariates? 17</p> <p>2.6.3 Are IPD Needed to Improve the Information Size? 17</p> <p>2.6.4 Are IPD Needed to Improve the Quality of Analysis? 18</p> <p>2.7 Concluding Remarks 19</p> <p><b>3 Planning and Initiating an IPD Meta-analysis Project </b><b>21<br /></b><i>Lesley A. Stewart, Richard D. Riley, and Jayne F. Tierney</i></p> <p>3.1 Introduction 22</p> <p>3.2 Organisational Approach 22</p> <p>3.2.1 Collaborative IPD Meta-analysis Project 22</p> <p>3.2.2 IPD Meta-analysis Projects Using Data Repositories or Data-sharing Platforms 24</p> <p>3.3 Developing a Project Scope 26</p> <p>3.4 Assessing Feasibility and ‘In Principle’ Support and Collaboration 26</p> <p>3.5 Establishing a Team with the Right Skills 29</p> <p>3.6 Advisory and Governance Functions 30</p> <p>3.7 Estimating How Long the Project Will Take 31</p> <p>3.8 Estimating the Resources Required 33</p> <p>3.9 Obtaining Funding 38</p> <p>3.10 Obtaining Ethical Approval 39</p> <p>3.11 Data-sharing Agreement 41</p> <p>3.12 Additional Planning for Prospective Meta-analysis Projects 41</p> <p>3.13 Concluding Remarks 43</p> <p><b>4 Running an IPD Meta-analysis Project: From Developing the Protocol to Preparing Data for Meta-analysis </b><b>45<br /></b><i>Jayne F. Tierney, Richard D. Riley, Larysa H.M. Rydzewska, and Lesley A. Stewart</i></p> <p>4.1 Introduction 46</p> <p>4.2 Preparing to Collect IPD 46</p> <p>4.2.1 Defining the Objectives and Eligibility Criteria 46</p> <p>4.2.2 Developing the Protocol for an IPD Meta-analysis Project 49</p> <p>4.2.3 Identifying and Screening Potentially Eligible Trials 51</p> <p>4.2.4 Deciding Which Information Is Needed to Summarise Trial Characteristics 51</p> <p>4.2.5 Deciding How Much IPD Are Needed 52</p> <p>4.2.6 Deciding Which Variables Are Needed in the IPD 52</p> <p>4.2.7 Developing a Data Dictionary for the IPD 55</p> <p>4.3 Initiating and Maintaining Collaboration 57</p> <p>4.4 Obtaining IPD 59</p> <p>4.4.1 Ensuring That IPD Are De-identified 59</p> <p>4.4.2 Providing Data Transfer Guidance 60</p> <p>4.4.3 Transferring trial IPD securely 61</p> <p>4.4.4 Storing Trial IPD Securely 61</p> <p>4.4.5 Making Best Use of IPD from Repositories 61</p> <p>4.5 Checking and Harmonising Incoming IPD 62</p> <p>4.5.1 The Process and Principles 63</p> <p>4.5.2 Initial Checking of IPD for Each Trial 63</p> <p>4.5.3 Harmonising IPD across Trials 64</p> <p>4.5.4 Checking the Validity, Range and Consistency of Variables 65</p> <p>4.6 Checking the IPD to Inform Risk of Bias Assessments 66</p> <p>4.6.1 The Randomisation Process 68</p> <p>4.6.2 Deviations from the Intended Interventions 71</p> <p>4.6.3 Missing Outcome Data 73</p> <p>4.6.4 Measurement of the Outcome 74</p> <p>4.7 Assessing and Presenting the Overall Quality of a Trial 76</p> <p>4.8 Verification of Finalised Trial IPD 77</p> <p>4.9 Merging IPD Ready for Meta-analysis 77</p> <p>4.10 Concluding Remarks 80</p> <p><b>Part I References </b><b>81</b></p> <p><b>Part II Fundamental Statistical Methods and Principles </b><b>87</b></p> <p><b>5 The Two-stage Approach to IPD Meta-analysis </b><b>89<br /></b><i>Richard D. Riley, Thomas P.A. Debray, Tim P. Morris, and Dan Jackson</i></p> <p>5.1 Introduction 90</p> <p>5.2 First Stage of a Two-stage IPD Meta-analysis 90</p> <p>5.2.1 General Format of Regression Models to Use in the First Stage 92</p> <p>5.2.2 Estimation of Regression Models Applied in the First Stage 92</p> <p>5.2.3 Regression for Different Outcome Types 94</p> <p>5.2.3.1 Continuous Outcomes 94</p> <p>5.2.3.2 Binary Outcomes 98</p> <p>5.2.3.3 Ordinal and Multinomial Outcomes 99</p> <p>5.2.3.4 Count and Incidence Rate Outcomes 100</p> <p>5.2.3.5 Time-to-Event Outcomes 101</p> <p>5.2.4 Adjustment for Prognostic Factors 102</p> <p>5.2.5 Dealing with Other Trial Designs and Missing Data 103</p> <p>5.3 Second Stage of a Two-stage IPD Meta-analysis 106</p> <p>5.3.1 Meta-analysis Assuming a Common Treatment Effect 106</p> <p>5.3.2 Meta-analysis Assuming Random Treatment Effects 107</p> <p>5.3.3 Forest Plots and Percentage Trial Weights 110</p> <p>5.3.4 Heterogeneity Measures and Statistics 110</p> <p>5.3.5 Alternative Weighting Schemes 112</p> <p>5.3.6 Frequentist Estimation of the Between-Trial Variance of Treatment Effect 113</p> <p>5.3.7 Deriving Confidence Intervals for the Summary Treatment Effect 113</p> <p>5.3.8 Bayesian Estimation Approaches 115</p> <p>5.3.8.1 An Introduction to Bayes’ Theorem and Bayesian Inference 115</p> <p>5.3.8.2 Using a Bayesian Meta-Analysis Model in the Second Stage 115</p> <p>5.3.8.3 Applied Example 117</p> <p>5.3.9 Interpretation of Summary Effects from Meta-analysis 118</p> <p>5.3.10 Prediction Interval for the Treatment Effect in a New Trial 118</p> <p>5.4 Meta-regression and Subgroup Analyses 120</p> <p>5.5 The <i>ipdmetan </i>Software Package 121</p> <p>5.6 Combining IPD with Aggregate Data from non-IPD Trials 124</p> <p>5.7 Concluding Remarks 125</p> <p><b>6 The One-stage Approach to IPD Meta-analysis </b><b>127<br /></b><i>Richard D. Riley and Thomas P.A. Debray </i>127</p> <p>6.1 Introduction 128</p> <p>6.2 One-stage IPD Meta-analysis Models Using Generalised Linear Mixed Models 129</p> <p>6.2.1 Basic Statistical Framework of One-stage Models Using GLMMs 129</p> <p>6.2.1.1 Continuous Outcomes 130</p> <p>6.2.1.2 Binary Outcomes 130</p> <p>6.2.1.3 Ordinal and Multinomial Outcomes 135</p> <p>6.2.1.4 Count and Incidence Rate Outcomes 136</p> <p>6.2.2 Specifying Parameters as Either Common, Stratified, or Random 136</p> <p>6.2.3 Accounting for Clustering of Participants within Trials 139</p> <p>6.2.3.1 Examples 141</p> <p>6.2.4 Choice of Stratified Intercept or Random Intercepts 141</p> <p>6.2.4.1 Findings from Simulation Studies 142</p> <p>6.2.4.2 Our Preference for Using a Stratified Intercept 142</p> <p>6.2.4.3 Allowing for Correlation between Random Effects on Intercept and Treatment Effect 143</p> <p>6.2.5 Stratified or Common Residual Variances 144</p> <p>6.2.6 Adjustment for Prognostic Factors 145</p> <p>6.2.7 Inclusion of Trial-level Covariates 145</p> <p>6.2.8 Estimation of One-stage IPD Meta-analysis Models Using GLMMs 146</p> <p>6.2.8.1 Software for Fitting One-stage Models 146</p> <p>6.2.8.2 ML Estimation and Downward Bias in Between-trial Variance Estimates 146</p> <p>6.2.8.3 Trial-specific Centering of Variables to Improve ML Estimation of One-stage Models with a Stratified Intercept 147</p> <p>6.2.8.4 REML Estimation 147</p> <p>6.2.8.5 Deriving Confidence Intervals for ParametersPpost-estimation 149</p> <p>6.2.8.6 Prediction Intervals 151</p> <p>6.2.8.7 Derivation of Percentage Trial Weights 151</p> <p>6.2.8.8 Bayesian Estimation for One-stage Models 151</p> <p>6.2.9 A Summary of Recommendations 152</p> <p>6.3 One-stage Models for Time-to-event Outcomes 152</p> <p>6.3.1 Cox Proportional Hazard Framework 152</p> <p>6.3.1.1 Stratifying Using Proportional Baseline Hazards and Frailty Models 152</p> <p>6.3.1.2 Stratifying Baseline Hazards without Assuming Proportionality 154</p> <p>6.3.1.3 Comparison of Approaches 154</p> <p>6.3.1.4 Estimation Methods 154</p> <p>6.3.1.5 Example 156</p> <p>6.3.2 Fully Parametric Approaches 157</p> <p>6.3.3 Extension to Time-varying Hazard Ratios and Joint Models 157</p> <p>6.4 One-stage Models Combining Different Sources of Evidence 159</p> <p>6.4.1 Combining IPD Trials with Partially Reconstructed IPD from Non-IPD Trials 159</p> <p>6.4.2 Combining IPD and Aggregate Data Using Hierarchical Related Regression 160</p> <p>6.4.3 Combining IPD from Parallel Group, Cluster and Cross-over Trials 161</p> <p>6.5 Reporting of One-stage Models in Protocols and Publications 162</p> <p>6.6 Concluding Remarks 162</p> <p><b>7 Using IPD Meta-analysis to Examine Interactions between Treatment Effect and Participant-level Covariates </b><b>163<br /></b><i>Richard D. Riley and David J. Fisher</i></p> <p>7.1 Introduction 164</p> <p>7.2 Meta-regression and Its Limitations 166</p> <p>7.2.1 Meta-regression of Aggregated Participant-level Covariates 166</p> <p>7.2.2 Low Power and Aggregation Bias 166</p> <p>7.2.3 Empirical Evidence of the Difference Between Using Across-trial and Within-trial Information to Estimate Treatment-covariate Interactions 167</p> <p>7.3 Two-stage IPD Meta-analysis to Estimate Treatment-covariate Interactions 168</p> <p>7.3.1 The Two-stage Approach 168</p> <p>7.3.2 Applied Example: Is the Effect of Anti-hypertensive Treatment Different for Males and Females? 170</p> <p>7.3.3 Do Not Quantify Interactions by Comparing Meta-analysis Results for Subgroups 171</p> <p>7.4 The One-stage Approach 174</p> <p>7.4.1 Merging Within-trial and Across-trial Information 174</p> <p>7.4.2 Separating Within-trial and Across-trial Information 175</p> <p>7.4.2.1 Approach (i) for a One-stage Survival Model: Center the Covariate and Include the Covariate Mean 175</p> <p>7.4.2.2 Approach (ii) for a One-stage Survival Model: Stratify All Nuisance Parameters by Trial 176</p> <p>7.4.2.3 Approaches (i) and (ii) for Continuous and Binary Outcomes 176</p> <p>7.4.2.4 Comparison of Approaches (i) and (ii) 177</p> <p>7.4.3 Applied Examples 177</p> <p>7.4.3.1 Is Age an Effect Modifier for Epilepsy Treatment? 177</p> <p>7.4.3.2 Is the Effect of an Early Support Hospital Discharge Modified by Having a Carer Present? 178</p> <p>7.4.4 Coding of the Treatment Covariate and Adjustment for Other Covariates 178</p> <p>7.4.4.1 Example 180</p> <p>7.4.5 Estimating the Aggregation Bias Directly 180</p> <p>7.4.6 Reporting Summary Treatment Effects for Subgroups after Adjusting for Aggregation Bias 180</p> <p>7.5 Combining IPD and non-IPD Trials 181</p> <p>7.5.1 Can We Recover Interaction Estimates from non-IPD Trials? 181</p> <p>7.5.2 How to Incorporate Interaction Estimates from non-IPD Trials in an IPD Metaanalysis 182</p> <p>7.6 Handling of Continuous Covariates 184</p> <p>7.6.1 Do Not Categorise Continuous Covariates 184</p> <p>7.6.2 Interactions May Be Non-linear 185</p> <p>7.6.2.1 Rationale and an Example 185</p> <p>7.6.2.2 Two-stage Multivariate IPD Meta-analysis for Summarising Non-linear Interactions 186</p> <p>7.6.2.3 One-stage IPD Meta-analysis for Summarising Non-linear Interactions 190</p> <p>7.7 Handling of Categorical or Ordinal Covariates 191</p> <p>7.8 Misconceptions and Cautions 191</p> <p>7.8.1 Genuine Treatment-covariate Interactions Are Rare 191</p> <p>7.8.2 Interactions May Depend on the Scale of Analysis 192</p> <p>7.8.3 Measurement Error May Impact Treatment-covariate Interactions 193</p> <p>7.8.4 Even without Treatment-covariate Interactions, the Treatment Effect on Absolute Risk May Differ across Participants 193</p> <p>7.8.5 Between-trial Heterogeneity in Treatment Effect Should Not Be Used to Guide Whether Treatment-covariate Interactions Exist at the Participant Level 194</p> <p>7.9 Is My Identified Treatment-covariate Interaction Genuine? 195</p> <p>7.10 Reporting of Analyses of Treatment-covariate Interactions 196</p> <p>7.11 Can We Predict a New Patient’s Treatment Effect? 196</p> <p>7.11.1 Linking Predictions to Clinical Decision Making 198</p> <p>7.12 Concluding Remarks 198</p> <p><b>8 One-stage versus Two-stage Approach to IPD Meta-analysis: Differences and Recommendations </b><b>199<br /></b><i>Richard D. Riley, Danielle L. Burke, and Tim Morris</i></p> <p>8.1 Introduction 200</p> <p>8.2 One-stage and Two-stage Approaches Usually Give Similar Results 200</p> <p>8.2.1 Evidence to Support Similarity of One-stage and Two-stage IPD Meta-analysis Results 200</p> <p>8.2.2 Examples 202</p> <p>8.2.3 Some Claims in Favour of the One-stage Approach Are Misleading 203</p> <p>8.3 Ten Key Reasons Why One-stage and Two-stage Approaches May Give Different Results 203</p> <p>8.3.1 Reason I: Exact One-stage Likelihood When Most Trials Are Small 204</p> <p>8.3.2 Reason II: How Clustering of Participants Within Trials Is Modelled 207</p> <p>8.3.3 Reason III: Coding of the Treatment Variable in One-stage Models Fitting with ML Estimation 208</p> <p>8.3.4 Reason IV: Different Estimation Methods for τ2 210</p> <p>8.3.5 Reason V: Specification of Prognostic Factor and Adjustment Terms 210</p> <p>8.3.6 Reason VI: Specification of the Residual Variances 212</p> <p>8.3.7 Reason VI: Choice of Common Effect or Random Effects for the Parameter of Interest 213</p> <p>8.3.8 Reason VIII: Derivation of Confidence Intervals 213</p> <p>8.3.9 Reason IX: Accounting for Correlation Amongst Multiple Outcomes or Time-points 214</p> <p>8.3.10 Reason X: Aggregation Bias for Treatment Covariate Interactions 215</p> <p>8.3.11 Other Potential Causes 215</p> <p>8.4 Recommendations and Guidance 216</p> <p>8.5 Concluding Remarks 217</p> <p><b>Part II References </b><b>219</b></p> <p><b>Part III Critical Appraisal and Dissemination </b><b>237</b></p> <p><b>9 Examining the Potential for Bias in IPD Meta-analysis Results </b><b>239<br /></b><i>Richard D. Riley, Jayne F. Tierney, and Lesley A. Stewart</i></p> <p>9.1 Introduction 240</p> <p>9.2 Publication and Reporting Biases of Trials 240</p> <p>9.2.1 Impact on IPD Meta-analysis Results 240</p> <p>9.2.2 Examining Small-study Effects Using Funnel Plots 241</p> <p>9.2.3 Small-study Effects May Arise Due to the Factors Causing Heterogeneity 243</p> <p>9.3 Biased Availability of the IPD from Trials 244</p> <p>9.3.1 Examining the Impact of Availability Bias 245</p> <p>9.3.2 Example: IPD Meta-analysis Examining High-dose Chemotherapy for the Treatment of Non-Hodgkin Lymphoma 246</p> <p>9.4 Trial Quality (risk of bias) 247</p> <p>9.5 Other Potential Biases Affecting IPD Meta-analysis Results 248</p> <p>9.5.1 Trial Selection Bias 248</p> <p>9.5.2 Selective Outcome Availability 250</p> <p>9.5.3 Use of Inappropriate Methods by the IPD Meta-analysis Research Team 250</p> <p>9.6 Concluding Remarks 251</p> <p><b>10 Reporting and Dissemination of IPD Meta-analyses </b><b>253<br /></b><i>Lesley A. Stewart, Richard D. Riley, and Jayne F. Tierney</i></p> <p>10.1 Introduction 253</p> <p>10.2 Reporting IPD Meta-analysis Projects in Academic Reports 254</p> <p>10.2.1 PRISMA-IPD Title and Abstract Sections 255</p> <p>10.2.2 PRISMA-IPD Introduction Section 259</p> <p>10.2.3 PRISMA-IPD Methods Section 259</p> <p>10.2.4 PRISMA-IPD Results Section 262</p> <p>10.2.5 PRISMA-IPD Discussion and Funding Sections 266</p> <p>10.3 Additional Means of Disseminating Findings 266</p> <p>10.3.1 Key Audiences 267</p> <p>10.3.1.1 The IPD Collaborative Group 267</p> <p>10.3.1.2 Patient and Public Audiences 267</p> <p>10.3.1.3 Guideline Developers 268</p> <p>10.3.2 Communication Channels 268</p> <p>10.3.2.1 Evidence Summaries and Policy Briefings 268</p> <p>10.3.2.2 Press Releases 268</p> <p>10.3.2.3 Social Media 270</p> <p>10.4 Concluding Remarks 270</p> <p><b>11 A Tool for the Critical Appraisal of IPD Meta-analysis Projects (CheckMAP) </b><b>271<br /></b><i>Jayne F. Tierney, Lesley A. Stewart, Claire L. Vale, and Richard D. Riley</i></p> <p>11.1 Introduction 271</p> <p>11.2 The CheckMAP Tool 272</p> <p>11.3 Was the IPD Meta-analysis Project Done within a Systematic Review Framework? 272</p> <p>11.4 Were the IPD Meta-analysis Project Methods Pre-specified in a Publicly Available Protocol? 274</p> <p>11.5 Did the IPD Meta-analysis Project Have a Clear Research Question Qualified by Explicit Eligibility Criteria? 276</p> <p>11.6 Did the IPD Meta-analysis Project Have a Systematic and Comprehensive Search Strategy? 276</p> <p>11.7 Was the Approach to Data Collection Consistent and Thorough? 277</p> <p>11.8 Were IPD Obtained from Most Eligible Trials and Their Participants? 277</p> <p>11.9 Was the Validity of the IPD Checked for Each Trial? 278</p> <p>11.10 Was the Risk of Bias Assessed for Each Trial and Its Associated IPD? 27811.10.1 Was the Randomisation Process Checked Based on IPD? 278</p> <p>11.10.2 Were the IPD Checked to Ensure That All (or Most) Randomised Participants Were Included? 279</p> <p>11.10.3 Were All Important Outcomes Included in the IPD? 279</p> <p>11.10.4 Were the Outcomes Measured/Defined Appropriately? 279</p> <p>11.10.5 Was the Quality of Outcome Data Checked? 280</p> <p>11.11 Were the Methods of Meta-analysis Appropriate? 280</p> <p>11.11.1 Were the Analyses Pre-specified in Detail and the Key Estimands Defined? 280</p> <p>11.11.2 Were the Methods of Summarising the Overall Effects of Treatments Appropriate? 281</p> <p>11.11.3 Were the Methods of Assessing whether Effects of Treatments Varied by Trial-level Characteristics Appropriate? 281</p> <p>11.11.4 Were the Methods of Assessing whether Effects of Treatments Varied by Participant-level Characteristics Appropriate? 282</p> <p>11.11.5 Was the Robustness of Conclusions Checked Using Relevant Sensitivity or Other Analyses? 282</p> <p>11.11.6 Did the IPD Meta-analysis Project’s Report Cover the Items Described in PRISMAIPD? 282</p> <p>11.12 Concluding Remarks 283</p> <p><b>Part III References </b><b>285</b></p> <p><b>Part IV Special Topics in Statistics </b><b>291</b></p> <p><b>12 Power Calculations for Planning an IPD Meta-analysis </b><b>293<br /></b><i>Richard D. Riley and Joie Ensor</i></p> <p>12.1 Introduction 294</p> <p>12.1.1 Rationale for Power Calculations in an IPD Meta-analysis 294</p> <p>12.1.2 Premise for This Chapter 294</p> <p>12.2 Motivating Example: Power of a Planned IPD Meta-analysis of Trials of Interventions to Reduce Weight Gain in Pregnant Women 295</p> <p>12.2.1 Background 295</p> <p>12.2.2 What Is the Power to Detect a Treatment-BMI Interaction? 295</p> <p>12.2.3 Power of an IPD Meta-analysis to Detect a Treatment-covariate Interaction for a Continuous Outcome 295</p> <p>12.2.4 Closed-form Solutions 296</p> <p>12.2.4.1 Application to the i-WIP Example 298</p> <p>12.2.5 Simulation-based Power Calculations for a Two-stage IPD Meta-analysis 299</p> <p>12.2.5.1 Application to the i-WIP Example 300</p> <p>12.2.6 Power Results Naively Assuming the IPD All Come from a Single Trial 301</p> <p>12.3 The Contribution of Individual Trials Toward Power 301</p> <p>12.3.1 Contribution According to Sample Size 301</p> <p>12.3.2 Contribution According to Covariate and Outcome Variability 302</p> <p>12.4 The Impact of Model Assumptions on Power 302</p> <p>12.4.1 Impact of Allowing for Heterogeneity in the Interaction 302</p> <p>12.4.2 Impact of Wrongly Modelling BMI as a Binary Variable 304</p> <p>12.4.3 Impact of Adjusting for Additional Covariates 304</p> <p>12.5 Extensions 305</p> <p>12.5.1 Power Calculations for Binary and Time-to-event Outcomes 305</p> <p>12.5.2 Simulation Using a One-stage IPD Meta-analysis Approach 306</p> <p>12.5.3 Examining the Potential Precision of IPD Meta-analysis Results 307</p> <p>12.5.4 Estimating the Power of a New Trial Conditional on IPD Meta-analysis Results 307</p> <p>12.6 Concluding Remarks 309</p> <p><b>13 Multivariate Meta-analysis Using IPD </b><b>311<br /></b><i>Richard D. Riley, Dan Jackson, and Ian R. White</i></p> <p>13.1 Introduction 312</p> <p>13.2 General Two-stage Approach for Multivariate IPD Meta-analysis 314</p> <p>13.2.1 First-stage Analyses 315</p> <p>13.2.1.1 Obtaining Treatment Effect Estimates and Their Variances for Continuous Outcomes 315</p> <p>13.2.1.2 Obtaining Within-trial Correlations Directly or via Bootstrapping for Continuous Outcomes 316</p> <p>13.2.1.3 Extension to Binary, Time-to-event and Mixed Outcomes 317</p> <p>13.2.2 Second-stage Analysis: Multivariate Meta-analysis Model 319</p> <p>13.2.2.1 Multivariate Model Structure 320</p> <p>13.2.2.2 Dealing with Missing Outcomes 320</p> <p>13.2.2.3 Frequentist Estimation of the Multivariate Model 321</p> <p>13.2.2.4 Bayesian Estimation of the Multivariate Model 322</p> <p>13.2.2.5 Joint Inferences and Predictions 322</p> <p>13.2.2.6 Alternative Specifications for the Between-trial Variance Matrix with Missing Outcomes 323</p> <p>13.2.2.7 Combining IPD and non-IPD Trials 323</p> <p>13.2.3 Useful Measures to Accompany Multivariate Meta-analysis Results 324</p> <p>13.2.3.1 Heterogeneity Measures 324</p> <p>13.2.3.2 Percentage Trial Weights 325</p> <p>13.2.3.3 The Efficiency (E) and Borrowing of Strength (BoS) Statistics 325</p> <p>13.2.4 Understanding the Impact of Correlation and Borrowing of Strength 326</p> <p>13.2.4.1 Anticipating the Value of BoS When Assuming Common Treatment Effects 326</p> <p>13.2.4.2 <i>BoS When Assuming Random Treatment Effects </i>327</p> <p>13.2.4.3 How the Borrowing of Strength Impacts upon the Summary Meta-analysis Estimates 327</p> <p>13.2.4.4 How the Correlation Impacts upon Joint Inferences across Outcomes 328</p> <p>13.2.5 Software 328</p> <p>13.3 Application to an IPD Meta-analysis of Anti-hypertensive Trials 329</p> <p>13.3.1 Bivariate Meta-analysis of SBP and DBP 329</p> <p>13.3.1.1 First-stage Results 329</p> <p>13.3.1.2 Second-stage Results 329</p> <p>13.3.1.3 Predictive Inferences 331</p> <p>13.3.2 Bivariate Meta-analysis of CVD and Stroke 332</p> <p>13.3.3 Multivariate Meta-analysis of SBP, DBP, CVD and Stroke 332</p> <p>13.4 Extension to Multivariate Meta-regression 333</p> <p>13.5 Potential Limitations of Multivariate Meta-analysis 334</p> <p>13.5.1 The Benefits of a Multivariate Meta-analysis for Each Outcome Are Often Small 335</p> <p>13.5.2 Model Specification and Estimation Is Non-trivial 335</p> <p>13.5.3 Benefits Arise under Assumptions 335</p> <p>13.6 One-stage Multivariate IPD Meta-analysis Applications 337</p> <p>13.6.1 Summary Treatment Effects 337</p> <p>13.6.1.1 Applied Example 337</p> <p>13.6.2 Multiple Treatment-covariate Interactions 337</p> <p>13.6.2.1 Applied Example 339</p> <p>13.6.3 Multinomial Outcomes 339</p> <p>13.7 Special Applications of Multivariate Meta-analysis 340</p> <p>13.7.1 Longitudinal Data and Multiple Time-points 340</p> <p>13.7.1.1 Applied Example 341</p> <p>13.7.1.2 Extensions 342</p> <p>13.7.2 Surrogate Outcomes 342</p> <p>13.7.3 Development of Multi-parameter Models for Dose Response and Prediction 344</p> <p>13.7.4 Test Accuracy 345</p> <p>13.7.5 Treatment-covariate Interactions 345</p> <p>13.7.5.1 Non-linear Trends 345</p> <p>13.7.5.2 Multiple Treatment-covariate Interactions 345</p> <p>13.8 Concluding Remarks 346</p> <p><b>14 Network Meta-analysis Using IPD </b><b>347<br /></b><i>Richard D. Riley, David M. Phillippo, and Sofia Dias</i></p> <p>14.1 Introduction 348</p> <p>14.2 Rationale and Assumptions for Network Meta-analysis 348</p> <p>14.3 Network Meta-analysis Models Assuming Consistency 350</p> <p>14.3.1 A Two-stage Approach 350</p> <p>14.3.2 A One-stage Approach 351</p> <p>14.3.3 Summary Results after a Network Meta-analysis 352</p> <p>14.3.4 Example: Comparison of Eight Thrombolytic Treatments after Acute Myocardial Infarction 352</p> <p>14.3.4.1 Two-stage Approach 353</p> <p>14.3.4.2 One-stage Approach 357</p> <p>14.4 Ranking Treatments 357</p> <p>14.5 How Do We Examine Inconsistency between Direct and Indirect Evidence? 359</p> <p>14.6 Benefits of IPD for Network Meta-analysis 361</p> <p>14.6.1 Benefit 1: Examining and Plotting Distributions of Covariates across Trials Providing Different Comparisons 361</p> <p>14.6.2 Benefit 2: Adjusting for Prognostic Factors to Improve Consistency and Reduce Heterogeneity 361</p> <p>14.6.3 Benefit 3: Including Treatment-covariate Interactions 362</p> <p>14.6.4 Benefit 4: Multiple Outcomes 365</p> <p>14.7 Combining IPD and Aggregate Data in Network Meta-analysis 365</p> <p>14.7.1 Multilevel Network Meta-regression 367</p> <p>14.7.2 Example: Treatments to Reduce Plaque Psoriasis 369</p> <p>14.8 Further Topics 370</p> <p>14.8.1 Accounting for Dose and Class 370</p> <p>14.8.2 Inclusion of ‘Real-world’ Evidence 372</p> <p>14.8.3 Cumulative Network Meta-analysis 372</p> <p>14.8.4 Quality Assessment and Reporting 372</p> <p>14.9 Concluding Remarks 372</p> <p><b>Part IV References </b><b>375</b></p> <p><b>Part V Diagnosis, Prognosis and Prediction </b><b>387</b></p> <p><b>15 IPD Meta-analysis for Test Accuracy Research </b><b>389<br /></b><i>Richard D. Riley, Brooke Levis, and Yemisi Takwoingi </i>389</p> <p>15.1 Introduction 390</p> <p>15.1.1 Meta-analysis of Test Accuracy Studies 390</p> <p>15.1.2 The Need for IPD 391</p> <p>15.1.3 Scope of This Chapter 394</p> <p>15.2 Motivating Example: Diagnosis of Fever in Children Using Ear Temperature 394</p> <p>15.3 Key Steps Involved in an IPD Meta-analysis of Test Accuracy Studies 397</p> <p>15.3.1 Defining the Research Objectives 397</p> <p>15.3.2 Searching for Studies with Eligible IPD 397</p> <p>15.3.3 Extracting Key Study Characteristics and Information 398</p> <p>15.3.4 Evaluating Risk of Bias of Eligible Studies 398</p> <p>15.3.5 Obtaining, Cleaning and Harmonising IPD 401</p> <p>15.3.6 Undertaking IPD Meta-analysis to Summarise Test Accuracy at a Particular Threshold 401</p> <p>15.3.6.1 Bivariate IPD Meta-analysis to Summarise Sensitivity and Specificity 401</p> <p>15.3.6.2 Examining and Summarising Heterogeneity 402</p> <p>15.3.6.3 Combining IPD and non-IPD Studies 403</p> <p>15.3.6.4 Application to the Fever Example 403</p> <p>15.3.6.5 Bivariate Meta-analysis of PPV and NPV 404</p> <p>15.3.7 Examining Accuracy-covariate Associations 406</p> <p>15.3.7.1 Model Specification Using IPD Studies 407</p> <p>15.3.7.2 Combining IPD and Aggregate Data 408</p> <p>15.3.7.3 Application to the Fever Example 408</p> <p>15.3.8 Performing Sensitivity Analyses and Examining Small-study Effects 409</p> <p>15.3.9 Reporting and Interpreting Results 409</p> <p>15.4 IPD Meta-analysis of Test Accuracy at Multiple Thresholds 410</p> <p>15.4.1 Separate Meta-analysis at Each Threshold 410</p> <p>15.4.2 Joint Meta-analysis of All Thresholds 410</p> <p>15.4.2.1 Modelling Using the Multinomial Distribution 411</p> <p>15.4.2.2 Modelling the Underlying Distribution of the Continuous Test Values 412</p> <p>15.5 IPD Meta-analysis for Examining a Test’s Clinical Utility 414</p> <p>15.5.1 Net Benefit and Decision Curves 415</p> <p>15.5.2 IPD Meta-analysis Models for Summarising Clinical Utility of a Test 416</p> <p>15.5.3 Application to the Fever Example 417</p> <p>15.6 Comparing Tests 418</p> <p>15.6.1 Comparative Test Accuracy Meta-analysis Models 419</p> <p>15.6.2 Applied Example 420</p> <p>15.7 Concluding Remarks 420</p> <p><b>16 IPD Meta-analysis for Prognostic Factor Research </b><b>421<br /></b><i>Richard D. Riley, Karel G.M. Moons, and Thomas P.A. Debray</i></p> <p>16.1 Introduction 422</p> <p>16.1.1 Problems with Meta-analyses Based on Published Aggregate Data 422</p> <p>16.1.2 Scope of This Chapter 424</p> <p>16.2 Potential Advantages of an IPD Meta-analysis 424</p> <p>16.2.1 Standardise Inclusion Criteria and Definitions 424</p> <p>16.2.2 Standardise Statistical Analyses 425</p> <p>16.2.3 Advanced Statistical Modelling 426</p> <p>16.3 Key Steps Involved in an IPD Meta-analysis of Prognostic Factor Studies 427</p> <p>16.3.1 Defining the Research Question 427</p> <p>16.3.1.1 Unadjusted or Adjusted Prognostic Factor Effects? 429</p> <p>16.3.2 Searching and Selecting Eligible Studies and Datasets 430</p> <p>16.3.3 Extracting Key Study Characteristics and Information 433</p> <p>16.3.4 Evaluating Risk of Bias of Eligible Studies 433</p> <p>16.3.5 Obtaining, Cleaning and Harmonising IPD 433</p> <p>16.3.6 Undertaking IPD Meta-analysis to Summarise Prognostic Effects 434</p> <p>16.3.6.1 A Two-stage Approach Assuming a Linear Prognostic Trend 434</p> <p>16.3.6.2 A Two-stage Approach with Non-linear Trends Using Splines or Polynomials 435</p> <p>16.3.6.3 Incorporating Measurement Error 438</p> <p>16.3.6.4 A One-stage Approach 440</p> <p>16.3.6.5 Checking the Proportional Hazards Assumption 441</p> <p>16.3.6.6 Dealing with Missing Data and Adjustment Factors 441</p> <p>16.3.7 Examining Heterogeneity and Performing Sensitivity Analyses 442</p> <p>16.3.8 Examining Small-study Effects 442</p> <p>16.3.9 Reporting and Interpreting Results 443</p> <p>16.4 Software 444</p> <p>16.5 Concluding Remarks 444</p> <p><b>17 IPD Meta-analysis for Clinical Prediction Model Research </b><b>447<br /></b><i>Richard D. Riley, Kym I.E. Snell, Laure Wynants, Valentijn M.T. de Jong, Karel G.M. Moons, and Thomas P.A. Debray</i></p> <p>17.1 Introduction 448</p> <p>17.2 IPD Meta-analysis for Prediction Model Research 448</p> <p>17.2.1 Types of Prediction Model Research 448</p> <p>17.2.2 Why IPD Meta-analyses Are Needed 450</p> <p>17.2.3 Key Steps Involved in an IPD Meta-analysis for Prediction Model Research 452</p> <p>17.2.3.1 Define the Research Question and PICOTS System 452</p> <p>17.2.3.2 Identify Relevant Existing Studies and Datasets 452</p> <p>17.2.3.3 Examine Eligibility and Risk of Bias of IPD 452</p> <p>17.2.3.4 Obtain, Harmonise and Summarise IPD 454</p> <p>17.2.3.5 Undertake Meta-analysis and Quantify Heterogeneity 455</p> <p>17.3 External Validation of an Existing Prediction Model Using IPD Meta-analysis 455</p> <p>17.3.1 Measures of Predictive Performance in a Single Study 456</p> <p>17.3.1.1 Overall Measures of Model Fit 456</p> <p>17.3.1.2 Calibration Plots and Measures 456</p> <p>17.3.1.3 Discrimination Measures 456</p> <p>17.3.2 Potential for Heterogeneity in a Model’s Predictive Performance 459</p> <p>17.3.2.1 Causes of Heterogeneity in Model Performance 460</p> <p>17.3.2.2 Disentangling Sources of Heterogeneity 461</p> <p>17.3.3 Statistical Methods for IPD Meta-analysis of Predictive Performance 461</p> <p>17.3.3.1 Two-stage IPD Meta-analysis 461</p> <p>17.3.3.2 Example 1: Validation of Prediction Models for Cardiovascular Disease 463</p> <p>17.3.3.3 Example 2: Meta-analysis of Case-mix Standardised Estimates of Model Performance 466</p> <p>17.3.3.4 Example 3: Examining Predictive Performance of QRISK2 across Multiple Practices 468</p> <p>17.3.3.5 One-stage IPD Meta-analysis 469</p> <p>17.4 Updating and Tailoring of a Prediction Model Using IPD Meta-analysis 470</p> <p>17.4.1 Example 1: Updating of the Baseline Hazard in a Prognostic Prediction Model 470</p> <p>17.4.2 Example 2: Multivariate IPD Meta-analysis to Compare Different Model Updating Strategies 471</p> <p>17.5 Comparison of Multiple Existing Prediction Models Using IPD Meta-analysis 472</p> <p>17.5.1 Example 1: Comparison of QRISK2 and Framingham 472</p> <p>17.5.2 Example 2: Comparison of Prediction Models for Pre-eclampsia 476</p> <p>17.5.3 Comparing Models When Predictors Are Unavailable in Some Studies 476</p> <p>17.6 Using IPD Meta-analysis to Examine the Added Value of a New Predictor to an Existing Prediction Model 478</p> <p>17.7 Developing a New Prediction Model Using IPD Meta-analysis 479</p> <p>17.7.1 Model Development Issues 479</p> <p>17.7.1.1 Examining and Handling Between-study Heterogeneity in Case-mix Distributions 479</p> <p>17.7.1.2 One-stage or Two-stage IPD Meta-analysis Models 482</p> <p>17.7.1.3 Allowing for Between-study Heterogeneity and Inclusion of Study-specific Parameters 483</p> <p>17.7.1.4 Studies with Different Designs 484</p> <p>17.7.1.5 Predictor Selection Based on Statistical Significance 484</p> <p>17.7.1.6 Conditional and Marginal Apparent Performance 485</p> <p>17.7.1.7 Sample Size, Overfitting and Penalisation 485</p> <p>17.7.2 Internal-external Cross-validation to Examine Transportability 487</p> <p>17.7.2.1 Overview of the Method 487</p> <p>17.7.2.2 Example: Diagnostic Prediction Model for Deep Vein Thrombosis 488</p> <p>17.8 Examining the Utility of a Prediction Model Using IPD Meta-analysis 491</p> <p>17.8.1 Example: Net Benefit of a Diagnostic Prediction Model for Ovarian Cancer 492</p> <p>17.8.1.1 Summary and Predicted Net Benefit of the LR2 Model 493</p> <p>17.8.1.2 Comparison to Strategies of Treat All or Treat None 493</p> <p>17.8.2 Decision Curves 493</p> <p>17.9 Software 494</p> <p>17.10 Reporting 495</p> <p>17.11 Concluding Remarks 495</p> <p><b>18 Dealing with Missing Data in an IPD Meta-analysis </b><b>499<br /></b><i>Thomas Debray, Kym I.E. Snell, Matteo Quartagno, Shahab Jolani, Karel G.M. Moons, and Richard D. Riley </i>499</p> <p>18.1 Introduction 500</p> <p>18.2 Motivating Example: IPD Meta-analysis Validating Prediction Models for Risk of Preeclampsia in Pregnancy 500</p> <p>18.3 Types of Missing Data in an IPD Meta-analysis 502</p> <p>18.4 Recovering Actual Values of Missing Data within IPD 502</p> <p>18.5 Mechanisms and Patterns of Missing Data in an IPD Meta-analysis 502</p> <p>18.5.1 Mechanisms of Missing Data 504</p> <p>18.5.2 Patterns of Missing Data 504</p> <p>18.5.3 Example: Risk of Pre-eclampsia in Pregnancy 505</p> <p>18.6 Multiple Imputation to Deal with Missing Data in a Single Study 506</p> <p>18.6.1 Joint Modelling 506</p> <p>18.6.2 Fully Conditional Specification 507</p> <p>18.6.3 How Many Imputations Are Required? 508</p> <p>18.6.4 Combining Results Obtained from Each Imputed Dataset 508</p> <p>18.7 Ensuring Congeniality of Imputation and Analysis Models 509</p> <p>18.8 Dealing with Sporadically Missing Data in an IPD Meta-analysis by Applying Multiple Imputation for Each Study Separately 509</p> <p>18.8.1 Example: Risk of Pre-eclampsia in Pregnancy 511</p> <p>18.9 Dealing with Systematically Missing Data in an IPD Meta-analysis Using a Bivariate Metaanalysis of Partially and Fully Adjusted Results 511</p> <p>18.10 Dealing with Both Sporadically and Systematically Missing Data in an IPD Meta-analysis Using Multilevel Modelling 514</p> <p>18.10.1 Motivating Example: Prognostic Factors for Short-term Mortality in Acute Heart Failure 515</p> <p>18.10.2 Multilevel Joint Modelling 516</p> <p>18.10.3 Multilevel Fully Conditional Specification 519</p> <p>18.11 Comparison of Methods and Recommendations 521</p> <p>18.11.1 Multilevel FCS versus Joint Model Approaches 521</p> <p>18.11.2 Sensitivity Analyses and Reporting 523</p> <p>18.12 Software 523</p> <p>18.13 Concluding Remarks 524</p> <p><b>Part V References </b><b>525</b></p> <p>Index 000</p>
<p><b>Richard D. Riley</b> is Professor of Biostatistics in the School of Medicine, Keele University, UK.</p><p><b>Jayne F. Tierney</b> is Professor of Evidence Synthesis at the MRC Clinical Trials Unit, University College London, UK.</p><p><b>Lesley A. Stewart</b> is Professor of Evidence Synthesis and Director of the Centre for Reviews and Dissemination, University of York, UK.</p>
<p><i>Individual Participant Data Meta-Analysis: A Handbook for Healthcare Research</i> provides a comprehensive introduction to the fundamental principles and methods that healthcare researchers need when considering, conducting or using individual participant data (IPD) meta-analysis projects. Written and edited by researchers with substantial experience in the field, the book details key concepts and practical guidance for each stage of an IPD meta-analysis project, alongside illustrated examples and summary learning points.</p><p>Split into five parts, the book chapters take the reader through the journey from initiating and planning IPD projects to obtaining, checking, and meta-analysing IPD, and appraising and reporting findings. The book initially focuses on the synthesis of IPD from randomised trials to evaluate treatment effects, including the evaluation of participant-level effect modifiers (treatment-covariate interactions). Detailed extension is then made to specialist topics such as diagnostic test accuracy, prognostic factors, risk prediction models, and advanced statistical topics such as multivariate and network meta-analysis, power calculations, and missing data.</p><p>Intended for a broad audience, the book will enable the reader to:</p><ul><li>Understand the advantages of the IPD approach and decide when it is needed over a conventional systematic review</li><li>Recognise the scope, resources and challenges of IPD meta-analysis projects</li><li>Appreciate the importance of a multi-disciplinary project team and close collaboration with the original study investigators</li><li>Understand how to obtain, check, manage and harmonise IPD from multiple studies</li><li>Examine risk of bias (quality) of IPD and minimise potential biases throughout the project</li><li>Understand fundamental statistical methods for IPD meta-analysis, including two-stage and one-stage approaches (and their differences), and statistical software to implement them</li><li>Clearly report and disseminate IPD meta-analyses to inform policy, practice and future research</li><li>Critically appraise existing IPD meta-analysis projects</li><li>Address specialist topics such as effect modification, multiple correlated outcomes, multiple treatment comparisons, non-linear relationships, test accuracy at multiple thresholds, multiple imputation, and developing and validating clinical prediction models</li></ul><p>Detailed examples and case studies are provided throughout.</p>

Diese Produkte könnten Sie auch interessieren:

Statistics for Microarrays
Statistics for Microarrays
von: Ernst Wit, John McClure
PDF ebook
90,99 €