Real-World Evidence in Medical Product Development

Description

This book provides state-of-art statistical methodologies, practical considerations from regulators and sponsors, logistics, and real use cases for practitioners for the uptake of RWE/D. Randomized clinical trials have been the gold standard for the evaluation of efficacy and safety of medical products. However, the cost, duration, practicality, and limited generalizability have incentivized many to look for alternative ways to optimize drug development. This book provides a comprehensive list of topics together to include all aspects with the uptake of RWE/D, including, but not limited to, applications in regulatory and non-regulatory settings, causal inference methodologies, organization and infrastructure considerations, logistic challenges, and practical use cases. Preface Contents Editors and Contributors About the Editors About the Contributors Part I Real-World Data and Evidence to Accelerate Medical Product Development The Need for Real-World Evidence in Medical Product Development and Future Directions 1 Introduction 2 Where We Are Now with the Use of RWE and RWD 2.1 Regulatory Advancement 2.2 Advancement in Operational Considerations 2.3 Advancement in Statistical Methodologies in Causal Inference 2.4 Advancement in Real Case Applications 3 Opportunities for Further Advancement 3.1 Regulatory Context 3.2 Clinical Context 3.3 Study Design and Analysis Context 3.4 Data Context 3.5 Governance and Infrastructure Context 4 Future Direction and Concluding Remarks References Overview of the Current Real-World Evidence Regulatory Landscape 1 Introduction 2 Key Concepts in Real-World Evidence Worldwide 2.1 Sources of Real-World Data and Real-World Evidence 2.2 Regulatory Acceptability and Demonstrating Fitness-for-Purpose 3 Regulatory Precedent Examples of Fit-for-Purpose Real-World Evidence 3.1 Scientific Purpose of Supporting Planning of Clinical Trials 3.2 Scientific Purpose of Supporting Safety and Effectiveness Evaluation 3.3 Scientific Purpose of Serving as External Control to a Clinical Study 3.4 Scientific Purpose of Supporting Extrapolating Efficacy or Safety Findings 4 Conclusions and Discussion References Key Considerations in Forming Research Questions and Conducting Research in Real-World Setting 1 Introduction 2 Gathering Knowledge 3 Forming Research Question 3.1 Population 3.2 Response/Outcome 3.3 Treatment/Exposure 3.4 Covariates (Counterfactual Thinking) 3.5 Time 4 Revising Research Question 5 Answering Research Question 6 Discussion References Part II Fit-for-Use RWD Assessment and Data Standards Assessment of Fit-for-Use Real-World Data Sources and Applications 1 Introduction 2 Gaining Insights on Aligning Research Questions with RWD Sources 2.1 Learning from RCT DUPLICATE Initiative 2.2 Further Insights on Aligning Research Question with RWD Sources 3 Semi-Quantitative Approach for Fit-for-Use RWD Assessment – Application of a Case Study 3.1 Estimand Related to Fit-for-Use RWD Assessment 3.2 Evaluation of Key Variables as Determined by Research Questions 3.3 Hypothetical Research Question and Quantitative Assessment Algorithms 3.4 Results 3.4.1 Relevancy 3.4.2 Reliability 4 Discussions and Conclusion References Key Variables Ascertainment and Validation in RW Setting 1 Introduction 2 Methods for Ascertainment 2.1 Rule-Based Methods 2.2 Machine Learning (ML)-Based Methods 2.3 Text Processing for Phenotyping 2.4 Ascertainment Through Linkage and Using Proxy 3 Validation 4 Special Consideration for Key Variables 4.1 Exposure 4.2 Outcome 4.3 Confounders 5 A Case Study from Myrbetriq® Postmarketing Requirement 6 Discussions and Concluding Remarks References Data Standards and Platform Interoperability 1 Why We Need to Scale Up the Generation and Use of Real-World Evidence 2 Enabling Health Information Interoperability 3 The Main Standards Used to Support Continuity of Health Care 3.1 Health Level Seven (HL7) 3.2 The International Organization for Standardization (ISO) 3.3 SNOMED CT 3.4 LOINC 3.5 The International Classification of Diseases (ICD) 3.6 DICOM 3.7 IHE 4 The Main Standards Used to Support Clinical Trials 4.1 CDISC 4.2 Federated Data Networks 4.3 What Is a Common Data Model, and Why Use One? 5 Making Data Fit for Shared Use 5.1 FAIR Principles 5.2 Data Quality 5.3 Research Infrastructures and Platforms 5.3.1 EHDEN 5.3.2 DARWIN EU® 5.3.3 European Health Data Space (EHDS) 6 Conclusion References Privacy-Preserving Record Linkage for Real-World Data 1 Introduction and Motivation 2 Data Preparation Methods 2.1 Data Preprocessing Methods 2.2 Privacy Protection Methods 2.2.1 Separation Principle 2.2.2 Secure Hash Encoding 2.2.3 Phonetic Encoding 2.2.4 Bloom Filters 3 Linkage Methods 3.1 Deterministic Linkage 3.2 Probabilistic Linkage 3.3 Unsupervised Classification Methods 4 Performance Evaluation 4.1 Measures 4.2 Assessment Method 5 Demonstration with the R Package RecordLinkage on Dataset NHANES 6 Discussion References Part III Causal Inference Framework and Methodologies in RWE Research Causal Inference with Targeted Learning for Producing and Evaluating Real-World Evidence 1 Introduction 2 Targeted Learning Estimation Roadmap 2.1 Step 0 2.2 Step 1 2.3 Step 2 2.4 Step 3 2.5 Step 4 2.5.1 Simulation Study 2.6 Step 5 3 Case Study: Single-Arm Trial with External Controls 3.1 Apply the TL Estimation Roadmap 3.1.1 Step 0 3.1.2 Step 1 3.1.3 Steps 2 and 3 3.1.4 Step 4 3.1.5 Step 5 4 Conclusion A Appendix A.1 Simulation Study Data Generation Process A.2 Case Study Data Generation Process References Estimand in Real-World Evidence Study: From Frameworks to Application 1 Introduction 2 Frameworks Relevant to Real-World Estimands 2.1 The Estimand Framework in ICH E9(R1) 2.2 Target Trial Framework 2.3 Causal Inference Framework 2.4 Targeted Learning Framework 3 Examples of Estimands in Real-World Evidence Studies 3.1 Single-Arm Trial with External Control 3.2 Longitudinal Study with a Static Treatment Regime 3.3 Longitudinal Study with a Dynamic Treatment Regime 4 Summary and Discussion References Clinical Studies Leveraging Real-World Data Using Propensity Score-based Methods 1 Introduction 2 Propensity Score and Type 1 Hybrid Studies 2.1 The Concept of Propensity Score 2.2 Estimation of Propensity Score and Assessment of Balance 2.3 The Two-Stage Paradigm for Study Design 2.4 An Illustrative Numerical Example of a Type 1 Hybrid Study 3 The Design and Analysis of Type 2 Hybrid Studies 3.1 Definition and Fundamental Statistical Issues 3.2 Using Power Prior or Composite Likelihood to Down-Weight RWD Patients 3.3 The Propensity Score Redefined 3.4 The Propensity Score-Integrated Approach for Type 2 Hybrid Studies 3.5 More Information on Outcome Analysis 4 The Design and Analysis of Type 3 Hybrid Studies 4.1 Definition and Fundamental Statistical Issues 4.2 The Balancing Property of Propensity Score in Type 3 Hybrid Studies 4.3 The Propensity Score-Integrated Approach for Type 3 Hybrid Studies 4.4 More Information on Outcome Analysis 4.5 Discussion References Recent Statistical Development for Comparative Effectiveness Research Beyond Propensity-Score Methods 1 Introduction 2 Conditional or Marginal 2.1 Propensity-Score Methods 2.2 Marginal Structural Models 3 Weighting or Standardization 3.1 The Weighting Strategy 3.1.1 Estimand 3.1.2 Initial Estimator 3.1.3 Doubly Robust Estimator 3.2 The Standardization Strategy 3.2.1 Estimand 3.2.2 Initial Estimator 3.2.3 Doubly Robust Estimator 3.3 Implementation and Comparison 4 Time-Independent or Time-Dependent 4.1 The Weighting Strategy 4.1.1 Estimand 4.1.2 Initial Estimator 4.1.3 Double-Robust Estimator 4.2 The Standardization Strategy 4.2.1 Estimand 4.2.2 Initial Estimator 4.2.3 Double-Robust Estimator 4.3 Implementation and Comparison 5 Discussion 5.1 Hypothetical Strategy for ICEs in Estimand Definition 5.2 Treatment-Policy Strategy for ICEs in Estimand Definition 5.3 Composite-Variable Strategy for ICEs in Estimand Definition 5.4 While-on-treatment Strategy for ICEs in Estimand Definition 5.5 Principal-Stratum Strategy for ICEs in Estimand Definition References Innovative Hybrid Designs and Analytical Approaches Leveraging Real-World Data and Clinical Trial Data 1 Introduction 2 Hybrid Designs and Analytical Approaches Leveraging Real-World External Controls and Clinical Trial Data 2.1 An Overview of Approaches for Leveraging External Control Data to Support Drug Development 2.2 Adaptive Designs That Mitigate Uncertainty About the Relevance of External Controls 2.2.1 Adaptive Approaches to Determining the Sample Size of an RCT with a Hybrid Control Arm 2.2.2 Adaptive Clinical Trial Designs Mitigating the Risk of an External Control Arm 2.3 Hybrid Adaptive Clinical Trials Using External Controls to Support Interim Decision-Making 2.4 Analytical Approaches for Combining External Controls and Clinical Trial Data 2.4.1 Comparing Bayesian Dynamic Borrowing and Propensity Score Analytic Approaches 2.4.2 Combining PS Matching and Bayesian Dynamic Borrowing 2.4.3 Combining PS Stratification and Bayesian Dynamic Borrowing 2.4.4 Combining PS Weighting or Parametric g-Estimation with the Bayesian Meta-analytic Approach 3 Randomized Controlled Studies Incorporating Real-World Data 3.1 Pragmatic Randomized Designs and Decentralized Randomized Designs 3.2 Scientific Considerations with PCT and DCT Hybrid Designs 3.2.1 Real-World Considerations, the Scientific Question/Estimand, and the Study Hypotheses 3.2.2 Real-World Endpoints and Statistical Methods Used to Support Validity and Fitness-for-Purpose 4 Discussion References Statistical Challenges for Causal Inference Using Time-to-Event Real-World Data 1 Introduction 2 Causal Estimands, Confounding Bias, and Population Adjustment When Using RWD 3 Adjustments for Causal Inference 4 The Selection of Time Zero 5 Pseudo-observations: An Approach for Easy Use of Complex Causal Inference Methods 6 Bayesian Approaches for Indirect Comparisons and Augmenting an Internal Control Arm 7 On the Use of Aggregated RWD 8 Other Topics 9 Summary and Areas of Further Researches References Sensitivity Analyses for Unmeasured Confounding: This Is the Way 1 Introduction 2 Causal Inference and Key Assumptions 3 Current State 3.1 Some Notation 4 Methods for Unmeasured Confounding Sensitivity Analyses 5 Advances in Broadly Applicable Methods 6 Proposed Best Practice 7 Conclusions References Sensitivity Analysis in the Analysis of Real-World Data 1 Introduction 2 Sensitivity Analysis of Identifiability Assumptions 2.1 Sensitivity Analysis of the Consistency Assumption 2.2 Sensitivity Analysis of the NUC Assumption 2.3 Sensitivity Analysis of the Positivity Assumption 3 Sensitivity Analysis of ICE Assumptions 3.1 Sensitivity Analysis for the Hypothetical Strategy 3.2 Sensitivity Analysis for the Treatment Policy Strategy 3.3 Sensitivity Analysis for the Composite Variable Strategy 3.4 Sensitivity Analysis for the While on Treatment Strategy 3.5 Sensitivity Analysis for the Principal Stratum Strategy 4 Sensitivity Analysis of Statistical Assumptions 5 Discussion References Personalized Medicine with Advanced Analytics 1 Background 1.1 What Is Personalized Medicine 1.2 Why Personalized Medicine 1.3 How to Practice Personalized Medicine 2 Role of Causal Inference and Advanced Analytics 2.1 Conditional Average Treatment Effects 2.2 Data Source for Personalized Medicine 3 Subgroup Analysis 3.1 Methods for Subgroup Identification 3.1.1 Identify a Subgroup by Thresholding Treatment Effect 3.1.2 Identify a Subgroup by Maximizing Difference of Treatment Effect Using Tree-Based Method 3.1.3 Identify a Subgroup by Selection of an Optimal Treatment Regime 3.2 Discussion 4 Dynamic Treatment Regime 4.1 Basic Framework 4.2 Methods for Estimating Optimal Dynamic Treatment Regimes 4.2.1 Indirect Estimation Methods 4.2.2 Direct Estimation Methods 4.3 Discussion 4.3.1 Alternative Outcome Types 4.3.2 Personalized Dose Finding 5 Conclusions References Use of Real-World Evidence in Health Technology Assessment Submissions 1 Introduction 2 Role of RWE in HTA Submissions 2.1 Data Sources and Types of RWE 2.2 Acceptability of RWE Across HTA Agencies 2.3 Role of RWE in Market Access and Reimbursement 3 Value and Strength of RWE for HTA Purposes 3.1 Efficacy–Effectiveness Gap and Strength of RWE 3.2 Case Studies 4 Guidelines for Use of RWE in HTA and Collaborative RWE Standard Development 4.1 NICE's New RWE Framework 4.2 ICER's 2020–2023 Value Assessment Framework 4.3 REALISE Guidance 4.4 HAS's Methodology Guidance 4.5 Collaboration Between CADTH and Health Canada 5 Discussion References Part IV Application and Case Studies Examples of Applying Causal-Inference Roadmap to Real-World Studies 1 Introduction 2 Cohort Studies with Continuous or Binary Outcomes 2.1 Describe the Observed Data and the Data Generating Experiment 2.2 Specify a Realistic Model for the Observed Data 2.3 Define the Target Estimand 2.4 Propose an Estimator of the Target Estimand 2.5 Obtain Estimate, Uncertainty Measurement, and Inference 2.6 Conduct Sensitivity Analysis and Interpret the Results 3 Single-arm Studies with External Controls 3.1 Describe the Observed Data and the Data Generating Experiment 3.2 Specify a Realistic Model for the Observed Data 3.3 Define the Target Estimand 3.4 Propose an Estimator of the Target Estimand 3.5 Obtain Estimate, Uncertainty Measurement, and Inference 3.6 Conduct Sensitivity Analysis and Interpret the Results 4 Cohort Studies with Intercurrent Events 4.1 Example of Using Hypothetical Strategy 4.1.1 Describe the Observed Data and the Data Generating Experiment 4.1.2 Specify a Realistic Model for the Observed Data 4.1.3 Define the Target Estimand 4.1.4 Propose an Estimator of the Target Estimand 4.1.5 Obtain Estimate, Uncertainty Measurement, and Inference 4.1.6 Conduct Sensitivity Analysis and Interpret the Results 4.2 Example of Using Treatment Policy Strategy 4.2.1 Describe the Observed Data and Data Generating Experiment 4.2.2 Specify a Realistic Model for the Observed Data 4.2.3 Define the Target Estimand 4.2.4 Propose an Estimator of the Target Estimand 4.2.5 Obtain Estimate, Uncertainty Measurement, and Inference 4.2.6 Conduct Sensitivity Analysis and Interpret the Results 4.3 Example of Using Composite Variable Strategy 4.3.1 Describe the Observed Data and the Data Generating Experiment 4.3.2 Specify a Realistic Model for the Observed Data 4.3.3 Define the Target Estimand 4.3.4 Propose an Estimator of the Target Estimand 4.3.5 Obtain Estimate, Uncertainty Measurement, and Inference 4.3.6 Conduct Sensitivity Analysis and Interpret the Results 4.4 Example of Using While on Treatment Strategy 4.4.1 Describe the Observed Data and Data Generating Experiment 4.4.2 Specify a Realistic Model for the Observed Data 4.4.3 Define the Target Estimand 4.4.4 Propose an Estimator of the Target Estimand 4.4.5 Obtain Estimate, Uncertainty Measurement, and Inference 4.4.6 Conduct Sensitivity Analysis and Interpret the Results 4.5 Example of Using Principal Stratum Strategy 4.5.1 Describe the Observed Data and the Data Generating Experiment 4.5.2 Specify a Realistic Model for the Observed Data 4.5.3 Define the Target Estimand 4.5.4 Propose an Estimator of the Target Estimand 4.5.5 Obtain Estimate, Uncertainty Measurement, and Inference 4.5.6 Conduct Sensitivity Analysis and Interpret the Results 5 Summary References Applications Using Real-World Evidence to Accelerate Medical Product Development 1 Introduction 2 RWE/RWD Case Studies by Regulatory Purposes 2.1 RWE/RWD as Part of the Original Marketing Application 2.1.1 Avelumab 2.1.2 Tafasitamab 2.2 RWE/RWD as Primary Data Source for Label Expansion 2.2.1 Prograf 2.2.2 SurgiMend 2.3 RWE/RWD as One of the Data Sources for Label Expansion 2.3.1 Orencia 2.4 RWE/RWD as Supplemental Information for the Regulatory Decision 2.4.1 Ibrance 3 Analysis of Key Considerations in the Regulatory Decisions 3.1 RWE/RWD Supporting the Original Marketing Application 3.1.1 Avelumab 3.1.2 Tafasitamab 3.2 RWE/RWD as the Primary Data Source for Label Expansion 3.2.1 Prograf 3.2.2 SurgiMend 3.3 RWE/RWD as One of the Data Sources for Label Expansion 3.3.1 Orencia 3.4 RWE/RWD as Supplemental Information for the Regulatory Decision 3.4.1 Ibrance 4 Lessons Learned and Best Practices 5 Conclusions References The Use of Real-World Data to Support the Assessment of the Benefit and Risk of a Medicine to Treat Spinal Muscular Atrophy 1 Introduction 1.1 Spinal Muscular Atrophy 1.2 Risdiplam 2 FIREFISH Study: External Control Data from Publications 2.1 Design and Methods 2.1.1 Study Design 2.1.2 Statistical Methodology: Performance Criteria Approach 2.2 Results 3 SUNFISH Study: External Control Data from Individual Patient Data 3.1 Design and Methods 3.1.1 Study Design 3.1.2 Statistical Methodology 3.2 Results 4 Discussion 4.1 Benefits of Using RWD 4.2 Challenges 4.3 Lessons Learned 5 Conclusion References Index