Textbook of Clinical Trials in Oncology. A Statistical Perspective

Cover
Half Title
Title Page
Copyright Page
Dedication
Contents
Acknowledgment
Editors
Contributors
1. Introduction to Clinical Trials
	1.1 Scope and Motivation
	1.2 Resources
	1.3 Conclusion
	References
Section I: Early to Middle Development
	2. Selection of Endpoints
		2.1 Introduction
		2.2 Key Definitions and Endpoint Selection
		2.3 Patient-Centered Endpoints
			2.3.1 Overall Survival
			2.3.2 Adverse Events and Toxicity
				2.3.2.1 Dose-Limiting Toxicity
			2.3.3 Health-Related Quality of Life
				2.3.3.1 European Organization for Research and Treatment of Cancer Quality-of-Life Questionnaire Core 30 Items
				2.3.3.2 Functional Assessment of Cancer Therapy – General Version
				2.3.3.3 Short-Form 36 Survey
		2.4 Tumor-Centered Endpoints
			2.4.1 Assessment of Response in Tumor-Centered Endpoints
			2.4.2 Progression-Free Survival and Time to Progression
			2.4.3 Disease-Free Survival
			2.4.4 Time to Treatment Failure
			2.4.5 Objective Response Rate and Duration of Response
		2.5 Endpoints under Evaluation
			2.5.1 Pathologic Complete Response (pCR)
			2.5.2 Immune-Related Response Criteria (irRC)
		References
	3. Innovative Phase I Trials
		3.1 Early-Phase Designs for Cytotoxic Agents
			3.1.1 Designs Based on Safety Endpoints
				3.1.1.1 Rule-Based Algorithms: “A + B” Designs
				3.1.1.2 Dose-Expansion Cohorts (DECs)
				3.1.1.3 Model-Based Designs
			3.1.2 Designs Based on Safety and Efficacy Endpoints
		3.2 Early-Phase Designs: Moving Beyond Cytotoxic Agents
			3.2.1 The Bayesian Quasi-CRM for Continuous Toxicity Endpoints
				3.2.1.1 Illustrative Example of Modeling Toxicity Scores: Quasi-CRM versus Conventional CRM
			3.2.2 Novel Endpoints in Early-Phase Trials
				3.2.2.1 Dose-Finding Designs Incorporating Pharmacokinetics (PK) Measures
				3.2.2.2 Dose-Finding Designs for Immunotherapies
		3.3 Conclusion
		References
	4. Current Issues in Phase II Cancer Clinical Trials
		4.1 Introduction
		4.2 Single-Arm Phase II Trials
			4.2.1 Optimal Two-Stage Designs
			4.2.2 Estimation of Response Rate
			4.2.3 Confidence Interval
			4.2.4 P-Value Calculation
		4.3 Phase II Trials with Heterogeneous Patient Populations
			4.3.1 Single-Stage Designs
			4.3.2 Example 4.6
			4.3.3 Two-Stage Designs
			4.3.4 Example 4.7
			4.3.5 Conditional P-Value
		4.4 Randomized Phase II Trials
			4.4.1 Single-Stage Design
			4.4.2 Two-Stage Design
				4.4.2.1 Choice of a1 and a2
				4.4.2.2 Choice of n1 and n2
			4.4.3 Numerical Studies
		4.5 Conclusion
		References
	5. Design and Analysis of Immunotherapy Clinical Trials
		5.1 Introduction
		5.2 Immune-Related Toxicity
		5.3 Delayed Treatment Benefit
		5.4 Marker Stratification
		5.5 Treatment Benefit in a Subset of Patients
		5.6 Conclusion
		Acknowledgment
		References
	6. Adaptive Designs
		6.1 Introduction
		6.2 Adaptive Designs for Dose-Finding Studies
		6.3 Population Finding
		6.4 Response-Adaptive Randomization
		6.5 Sample Size Re-Estimation
		6.6 Adaptive Seamless Designs
		6.7 Conclusion
		References
Section II: Late Phase Clinical Trials
	7. Sample Size Calculations for Phase III Trials in Oncology
		7.1 Introduction
		7.2 Basics of Sample Size Calculation in Phase III Oncology Trials
			7.2.1 Required Parameters and Settings
			7.2.2 Relationships among Survival Parameters
			7.2.3 Basic Parameters:
			7.2.4 Sample Size Calculations Using Additional Parameters
			7.2.5 Sample Size Calculations Based on the Log-Rank Test
		7.3 Software for Sample Size Calculations
		7.4 Superiority Trials
			7.4.1 Purpose of Superiority Trials
			7.4.2 The Sample Size Calculation Methods Used in Various Software Programs
				7.4.2.1 SAS Power Procedure: TWOSAMPLESURVIVAL Statement
				7.4.2.2 PASS: Log-Rank Tests and Tests for Two Survival Curves Using Cox’s Proportional Hazards Model
				7.4.2.3 SWOG Statistical Tool: Two-Arm Survival
			7.4.3 Example of a Superiority Trial (the EAGLE Trial)
			7.4.4 Comparison of the Sample Size Calculated with Each Software Program
				7.4.4.1 SAS Power Procedure
				7.4.4.2 PASS: Log-Rank Tests (Input Median Survival Times)
				7.4.4.3 PASS: Tests for Two Survival Curves Using Cox’s Proportional Hazards Model
				7.4.4.4 SWOG Statistical Tool Website
				7.4.4.5 Interpretation of the Results
		7.5 Non-Inferiority Trials
			7.5.1 Purpose of Non-Inferiority Trials and Formulas to Calculate the Sample Size
			7.5.2 Specification of the Non-Inferiority Margin,
			7.5.3 The Sample Size-Calculation Methods Used in Each Software Program
				7.5.3.1 SAS
				7.5.3.2 PASS: Non-Inferiority Log-Rank Tests and Tests for Two Survival Curves Using Cox’s Proportional Hazards Model
				7.5.3.3 SWOG Statistical Tool: Two-Arm Survival
			7.5.4 Example Trial (JCOG0404 Trial)
			7.5.5 Comparison of Sample Sizes Calculated with Each Software Program
				7.5.5.1 SAS Power Procedure
				7.5.5.2 PASS: Non-Inferiority Log-Rank Tests
				7.5.5.3 PASS: Non-Inferiority Tests for Two Survival Curves  Using Cox’s Proportional Hazards Model
				7.5.5.4 SWOG Statistical Tool Website
			7.5.6 Interpretation of the Results
		7.6 Other
			7.6.1 Consideration for One-Sided or Two-Sided Tests
			7.6.2 Violation of the Proportional-Hazards and Exponential-Curve Assumptions
		7.7 Conclusion
		References
	8. Non-Inferiority Trial
		8.1 Introduction
		8.2 Assumptions for NI Trials
			8.2.1 The Constancy of the Control Effect
			8.2.2 Assay Sensitivity
		8.3 Design
			8.3.1 Selecting the Active Control
			8.3.2 Determining the NI Margin
			8.3.3 Statistical Algorithm for Assessing Non-Inferiority
				8.3.3.1 The Fixed-Margin Approach
				8.3.3.2 Synthesis Approach
			8.3.4 Sample Size
			8.3.5 Other Design Alternatives and Issues
				8.3.5.1 Three-Arm Studies
				8.3.5.2 Switching between NI and Superiority
				8.3.5.3 Interim Analyses
		8.4 Trial Conduction
		8.5 Analyses
			8.5.1 Analysis Populations
			8.5.2 Missing Data
			8.5.3 NI and Superiority
		8.6 Reporting
		8.7 Examples
		References
	9. Design of Multi-Arm,  Multi-Stage Trials in Oncology
		9.1 Introduction
		9.2 Notation
			9.2.1 Multi-Arm Trial
			9.2.2 Multi-Arm, Multi-Stage
		9.3 Determining Statistical Quantities for Multi-Arm Trials
			9.3.1 Distribution of Test Statistics from a Multi-Arm Trial
				9.3.1.1 Normal Outcomes
				9.3.1.2 Binary Outcome
				9.3.1.3 Time-to-Event Outcome
			9.3.2 Evaluating the Operating Characteristics of a Multi-Arm Design
				9.3.2.1 Type I Error Rate
			9.3.3 Power
				9.3.3.1 Conjunctive Power
				9.3.3.2 Disjunctive Power
				9.3.3.3 Least Favorable Configuration
				9.3.3.4 Comparison of Power
			9.3.4 Case Study
		9.4 Designing Multi-Arm Multi-Stage Trials
			9.4.1 Distribution of Test Statistics
			9.4.2 Group-Sequential MAMS
				9.4.2.1 Example
				9.4.2.2 Extensions
			9.4.3 Drop-the-Loser Multi-Arm Trials
				9.4.3.1 Notation and Operating Characteristics
				9.4.3.2 Extensions
			9.4.4 Case Study
		9.5 Conclusion
		References
	10. Multiple Comparisons, Multiple Primary Endpoints and Subpopulation Analysis
		10.1 Sources of Multiplicity in Oncology Trials
			10.1.1 Introductory Example
		10.2 Multiple Testing Procedures
			10.2.1 Basic Concepts
				10.2.1.1 Error Rate in Confirmatory Clinical Trials
				10.2.1.2 Single-Step and Stepwise Procedures
				10.2.1.3 Closed Testing Procedures
				10.2.1.4 Adjusted Critical Values and Adjusted p-Values
				10.2.1.5 Simultaneous Confidence Intervals
			10.2.2 Common Multiple Testing Procedures
				10.2.2.1 Bonferroni Test
				10.2.2.2 Holm Procedure
				10.2.2.3 Hochberg Procedure
				10.2.2.4 Numerical Illustration
			10.2.3 Gatekeeping and Graphical Procedures Based on the CTP
				10.2.3.1 Bonferroni-Based Graphical Procedures
				10.2.3.2 Procedures Based on Asymptotic Normality
			10.2.4 Multiplicity Adjustment for Other Types of Endpoints
		10.3 Multiple Comparison Procedures in Oncology
			10.3.1 The Scope of Multiplicity Adjustment
			10.3.2 Multiple Endpoints Complications in Group Sequential Designs
			10.3.3 Outlook on Future Developments
		10.4 Conclusion
		References
	11. Cluster Randomized Trials
		11.1 Introduction
		11.2 Randomization
			11.2.1 Matching and Stratification
			11.2.2 Constrained Randomization
			11.2.3 Minimization
		11.3 Analysis
			11.3.1 Continuous Outcomes
				11.3.1.1 Model
				11.3.1.2 Estimation and Inference
				11.3.1.3 Example
			11.3.2 Dichotomous Outcomes
				11.3.2.1 Cluster-Level Proportions Model
				11.3.2.2 Cluster-Level Log-Odds Model
				11.3.2.3 Estimation and Inference
				11.3.2.4 Example
			11.3.3 Other Analysis Methods
		11.4 Sample Size and Power
			11.4.1 Continuous Outcomes
				11.4.1.1 Power
				11.4.1.2 Sample Size: Number of Clusters
				11.4.1.3 Sample Size per Cluster
				11.4.1.4 Unequal ICCs in Treatment Arms
				11.4.1.5 Unequal Allocation
				11.4.1.6 Covariates
				11.4.1.7 Varying Cluster Sizes
				11.4.1.8 Matching and Stratification
			11.4.2 Dichotomous Outcomes
				11.4.2.1 Sample Size and Power
				11.4.2.2 Sample Size per Cluster
				11.4.2.3 Unequal ICCs in Treatment Arms
				11.4.2.4 Unequal Allocation
				11.4.2.5 Covariates
				11.4.2.6 Varying Cluster Sizes
		11.5 Additional Resources
			11.5.1 Resources for Other Designs
			11.5.2 Resources for Power and Sample Size Calculation
		References
	12. Statistical Monitoring of Safety and Efficacy
		12.1 Introduction
		12.2 Monitoring of Safety
			12.2.1 Introduction
				12.2.1.1 Planning for Safety Monitoring
				12.2.1.2 Safety Monitoring: Sponsor View (Masked, Treatment Groups Pooled)
				12.2.1.3 Safety Monitoring: Data Monitoring Committee View (Partially or Completely Unmasked)
		12.3 Efficacy and Futility Monitoring
			12.3.1 Introduction
			12.3.2 Superiority Monitoring
			12.3.3 Futility Monitoring
			12.3.4 Non-Inferiority Monitoring
		12.4 Adaptive Designs
			12.4.1 Sample Size Re-Estimation
			12.4.2 Adaptive Design Challenges to DMCs
			12.4.3 Master Protocol Designs
		12.5 Centralized Risk-Based Monitoring
		12.6 Conclusion
		References
Section III: Personalized Medicine
	13. Biomarker-Based Phase II and III Clinical Trials in Oncology
		13.1 Introduction
		13.2 Phase II Trials
			13.2.1 Single-Arm Trials
			13.2.2 Randomized Phase II Trials with a Control Arm
			13.2.3 Master Protocol
				13.2.3.1 Umbrella/Platform Trials
				13.2.3.2 Basket Trials
		13.3 Phase III Trials
			13.3.1 Enrichment Designs
			13.3.2 Marker-Based, All-Comers Designs
				13.3.2.1 Null Hypothesis
				13.3.2.2 Weak or Strong Control
				13.3.2.3 Statistical Analysis Plans in the Marker-Stratified Design
				13.3.2.4 Numerical Evaluations of Statistical Analysis Plans
				13.3.2.5 Interim Analysis
				13.3.2.6 Unstratified Trials with Adaptive Designs for Marker Development and Validation
			13.3.3 Strategy Designs
		13.4 Conclusion
		Appendix: Asymptotic Distributions of the Test Statistics
		Acknowledgments
		References
	14. Genomic Biomarker Clinical Trial Designs
		14.1 Introduction
		14.2 Phase II Designs
			14.2.1 Basket Designs
			14.2.2 Platform Designs
		14.3 Enrichment Design
			14.3.1 Umbrella Design
		14.4 Adaptive Enrichment Designs
			14.4.1 Adaptive Enrichment with Single Binary Covariate
			14.4.2 Adaptive Enrichment with a Quantitative Biomarker
		14.5 Conclusion
		References
	15. Trial Designs for Rare Diseases and Small Samples in Oncology
		15.1 Introduction
		15.2 Using External Information and Extrapolation for Planning and Conducting Early-Phase Clinical Trials: Case Study in Pediatric Oncology
			15.2.1 Specification of the Dose Range
			15.2.2 Specification of the Working Model
			15.2.3 Calibration of the Prior Distribution
			15.2.4 Conclusions
		15.3 A General Design for a Confirmatory Basket Trial
			15.3.1 Overview of General Confirmatory Basket-Trial Design
			15.3.2 Pruning, Random High Bias, and Type I Error Control
			15.3.3 An Application Example
			15.3.4 Conclusions
		15.4 Decision Theoretic Methods for Small Populations and Biomarker Trials
			15.4.1 Simultaneously Optimizing Trial Designs and Decision Rules: A Case Study in Small Populations
			15.4.2 Optimizing Trial Designs Only: A Case Study in the Development of Targeted Therapies
			15.4.3 Conclusion
		Acknowledgments
		Conflict of Interest Disclosure
		References
	16. Statistical Methods for Biomarker and Subgroup Evaluation in Oncology Trials
		16.1 Introduction
		16.2 Overview of Methods for Biomarker and Subgroup Evaluation
			16.2.1 Traditional Approaches to Biomarker and Subgroup Evaluation
			16.2.2 Taxonomy of Modern Approaches to Biomarker and Subgroup Evaluation
			16.2.3 Global Outcome Modeling Methods
			16.2.4 Global Treatment Effect Modeling with Interaction Trees
			16.2.5 Modeling Optimal Treatment Regimes
		16.3 Case Study
		16.4 SIDES Methodology for Subgroup Identification
			16.4.1 Subgroup Generation
			16.4.2 Basic Constrained Subgroup Search: Treatment-Effect Restrictions
			16.4.3 Constraints on the Search Space: Biomarker Screening via SIDEScreen
			16.4.4  Constraints on the Search Space: Biomarker Screening via Stochastic SIDEScreen
			16.4.5 Interpretation Tools: Subgroup Proximity Measures
			16.4.6 Interpretation Tools: Adjustment for Selection Bias
			16.4.7 Interpretation Tools: Honest Treatment Effect Estimates
		16.5 Conclusion
			16.5.1 Principles of Data-Driven Subgroup Analysis
			16.5.2 Application of Principled Subgroup Analysis Methods
			16.5.3 Relative Advantages of Principled Subgroup Analysis Methods
			16.5.4 Challenges and Extensions
		References
	17. Developing and Validating Prognostic Models of Clinical Outcomes
		17.1 Introduction
		17.2 Use of Prognostic Factors in Trials
		17.3 Design of Prognostic Studies
			17.3.1 Sample Size Justification
		17.4 Identification of Prognostic Factors
			17.4.1 Shrinkage Methods
			17.4.2 Prostate Cancer Example Predicting Overall Survival
			17.4.3 Non-Parametric Approaches
			17.4.4 Prostate Cancer Example Predicting PSA Decline
		17.5 Common Pitfalls with Modeling
		17.6 Constructing Risk Groups
		17.7 Validation and Assessment of Prognostic Models
			17.7.1 External Validation of Overall Survival Model
			17.7.2 External Validation Predicting PSA Decline
		17.8 High Dimensional Space, Low Sample Size
			17.8.1 Example of SNPs Identification in Ultra-High Dimension
		17.9 Conclusion
		Acknowledgments
		References
	18. High-Dimensional, Penalized-Regression Models in Time-to-Event Clinical Trials
		18.1 Development and Validation of Multimarker Signatures
		18.2 Penalized Regression for Survival Endpoints
			18.2.1 Statistical Framework
			18.2.2 Maximum Penalized Likelihood Inference
				18.2.2.1 Ridge Penalization
				18.2.2.2 Lasso Penalization
				18.2.2.3 Adaptive Lasso Penalization
				18.2.2.4 Group Lasso Penalization
			18.2.3 The Choice of the Shrinkage Parameter
				18.2.3.1 Cross-Validation
				18.2.3.2 One-Standard-Error Rule
				18.2.3.3 Percentile-Lasso
				18.2.3.4 Stability Selection
		18.3 Signatures Predicting a Survival Outcome
			18.3.1 False Positives
			18.3.2 Conservative Choices of the Shrinkage Parameter
			18.3.3 Breast Cancer Application
		18.4 Signatures Predicting the Treatment Benefit
			18.4.1 Identification of Treatment-Effect Modifiers
			18.4.2 Estimation of the Expected Survival Probabilities
			18.4.3 Breast Cancer Application
		18.5 R Software
			18.5.1 Biomarker Selection
			18.5.2 Diagnostics
			18.5.3 Expected Survival Estimation
		18.6 Conclusion
		Declarations
			Acknowledgments
			Funding
		References
	19. Sequential, Multiple Assignment, Randomized Trials
		19.1 Motivating Sequential, Multiple Assignment, Randomized Trials by Dynamic Treatment Regimens
		19.2 Introduction to Sequential, Multiple Assignment, Randomized Trials
		19.3 SMART Compared to Other Trial Designs
		19.4 Alternative Trial Strategies to a SMART
		19.5 Research Questions Addressed by SMARTs
		19.6 Sample Size Calculations
		19.7 Methods for Analysis
		19.8 Illustrations/Data Analysis Examples
		19.9 Conclusion
		References
Section IV: Advanced Topics
	20. Assessing the Value of Surrogate Endpoints
		20.1 Introduction
		20.2 Motivating Examples
			20.2.1 Top Trial: Pathologic Complete Response in Breast Cancer after Anthracyclines
			20.2.2 The GASTRIC Initiative: Disease-Free Survival, Progression-Free Survival, and Stomach Cancers
		20.3 Statistical Requirements to Validate Surrogate Endpoints from Single Trials
			20.3.1 What Is Not Sufficient: Patient-Level Correlation
			20.3.2 What Is Controversial: Evidence of No Residual Effect from a Single Trial
			20.3.3 What Is Controversial: Quantifying the Treatment Effect Explained by the Surrogate Endpoint from a Single Trial
			20.3.4 Area of Research: Causal Inference
		20.4 Statistical Methods and Software to Validate Surrogate Endpoints from Multiple Trials
			20.4.1 The Meta-Analytic Approach: Introduction to a Reference Method
			20.4.2 Statistical Implementation
			20.4.3 Survival Endpoints
				20.4.3.1 Copula Approach
				20.4.3.2 Simple Weighted Regression
				20.4.3.3 Joint Poisson Model
				20.4.3.4 GASTRIC Example
		20.5 Practical Difficulties with the Evaluation of Surrogate Endpoints
			20.5.1 Validation of Surrogate and Expected Gain
			20.5.2 Heterogeneity of Treatment Effects and Surrogacy Evaluation
			20.5.3 Extent of Application and New Class of Treatments
			20.5.4 Second-Line Treatments
		20.6 Conclusion
		References
	21. Competing Risks
		21.1 Introduction
		21.2 Notations and Quantities of Interest
		21.3 Inference and Joint Inference
			21.3.1 Regression Model
				21.3.1.1 Regression with Clustered Data
				21.3.1.2 Regression with Missing Cause of Failure
				21.3.1.3 Regression for Years Lost
		21.4 Tests
			21.4.1 Joint Tests
				21.4.1.1 Example 1: Hodgkin’s Disease
				21.4.1.2 Example 2: Bone Marrow Transplant (BMT) Study
			21.4.2 Joint Regression Analysis
				21.4.2.1  Example 1: Follicular Cell Lymphoma Study (Time to Progression and Progreesion-Free Survival)
				21.4.2.2 Example 2: BMT Study
			21.4.3 Sample Size
				21.4.3.1 Sample Size Calculation for Joint Inference of Cause-Specific Hazard and All-Causes Hazard
			21.4.4 The powerCompRisk. R Package
			21.4.5 An Example: The 4D Trial
		21.5 Conclusion
		References
	22. Cure Models in Cancer Clinical Trials
		22.1 Introduction
		22.2 Mixture Cure Models
			22.2.1 Model and Properties
			22.2.2 Interpretation
			22.2.3 Identifiability
			22.2.4 Model Estimation
			22.2.5 Model Implementation
		22.3 Promotion Time Cure Models
			22.3.1 Model and Properties
			22.3.2 Link with Other Models
			22.3.3 Interpretation
			22.3.4 Identifiability
			22.3.5 Model Estimation
			22.3.6 Model Implementation
			22.3.7 Extensions
		22.4 When to Use a Cure Model
			22.4.1 Presentation of the Simulations
			22.4.2 Simulations Results
		22.5 Melanoma Clinical Trial
		22.6 Conclusion
		References
	23. Interval Censoring
		23.1 Introduction
			23.1.1 The Estimation Issue for the Naively Converting Approach
		23.2 Non-Parametric and Semi-Parametric Approaches for Analyzing Interval-Censored Data
			23.2.1 Non-Parametric Maximum Likelihood Estimation
			23.2.2 Comparing Survival Functions
			23.2.3 Proportional Hazards Model with Interval Censoring
		23.3 Analyzing Interval-Censored Data by R
			23.3.1 Package ICsurv
			23.3.2 Package interval
		23.4 Conclusion
		References
	24. Methods for Analysis of Trials with Changes from Randomized Treatment
		24.1 Introduction
		24.2 Defining the Question
		24.3 The Impact of Treatment Switching
		24.4 Illustrative Dataset
		24.5 Simple Methods for Adjusting for Treatment Changes: Per-Protocol Analysis
		24.6 Complex Methods for Adjusting for Treatment Changes
			24.6.1 Inverse Probability of Censoring Weights
				24.6.1.1 Concept
				24.6.1.2 Method
				24.6.1.3 Refinements
			24.6.2 Two-Stage Estimation
				24.6.2.1 Concept
				24.6.2.2 Method
				24.6.2.3 Refinements
			24.6.3 Rank Preserving Structural Failure Time Model
				24.6.3.1 Concept
				24.6.3.2 Method
				24.6.3.3 Refinements
		24.7 Identifying Appropriate Methods
			24.7.1 Assessment of Methodological Assumptions
			24.7.2 Assessment of “Model Diagnostics”
			24.7.3 Results of Simulation Studies
			24.7.4 Validation of Results
		24.8 Extensions
		References
	25. The Analysis of Adverse Events in Randomized Clinical Trials
		25.1 The Two Workhorses Incidence Rate and Incidence Proportion and Their Connection
			25.1.1 The Case of Uncensored Data
			25.1.2 The Case of Censored Data
			25.1.3 Data Example
		25.2 Beyond Incidence Rates and the Constant Hazards Assumption: Nelson-Aalen, Aalen-Johansen, Kaplan–Meier
			25.2.1 Non-Parametric Estimation
			25.2.2 Data Example
		25.3 Comparison of Treatment Groups
			25.3.1 Cox Model, Log-Rank Test
			25.3.2 Data Example
		25.4 Recurrent Adverse Events
		25.5 Conclusion
		References
	26. Analysis of Quality of Life Outcomes in Oncology Trials
		26.1 Introduction
		26.2 Three Broad Approaches to Analyzing Repeated QOL Assessments
		26.3 Example Dataset: The AIM-HIGH Trial
		26.4 Summarizing Repeated QOL Assessments
		26.5 Time by Time Analysis
		26.6 Response Feature Analysis: The Use of Summary Measures
		26.7 The Area Under the Curve (AUC)
		26.8 Calculating the AUC
		26.9 Other Summary Measures
		26.10 Longitudinal Models
		26.11 Autocorrelation
		26.12 Analyzing Longitudinal Quality of Life Outcome Data with a Marginal Model
		26.13 Treatment × Time Interactions and Baseline Measurements
		26.14 Example of Using a Marginal Model to Analyze Longitudinal QOL Outcome Data: The AIM-HIGH Trial
		26.15 Checking the Assumptions
		26.16 Random-Effects Models
		26.17 Random Slopes
		26.18 Missing Data
		26.19 Types and Patterns of Missing Data
		26.20 Describing the Extent and Patterns of Missing Data
		26.21 What If We Think the Data Is MNAR?
		26.22 Random-Effects versus Marginal Models
		26.23 Conclusion
		References
	27. Missing Data
		27.1 Introduction
		27.2 Trial Design
		27.3 Missing Data Mechanisms
			27.3.1 Types of Missing Data Mechanisms
			27.3.2 Detecting Missing Data Mechanisms
		27.4 Analysis of Missing Data
			27.4.1 Missing Completely at Random
			27.4.2 Missing at Random
				27.4.2.1 Mixed Effects Models
				27.4.2.2 Imputation
			27.4.3 Missing Not at Random
				27.4.3.1 Pattern Mixture Models
				27.4.3.2 Shared Parameter Models
				27.4.3.3 Selection Models
				27.4.3.4 Multiple-Model Multiple Imputation
			27.4.4 Sensitivity Analyses
			27.4.5 Missing Baseline Covariates
		27.5 Conclusion
		References
Index