Design and Analysis of Subgroups with Biopharmaceutical Applications (Emerging Topics in Statistics and Biostatistics)

Preface
Contents
List of Contributors
Reviewers of the Subgroup Book
Part I Subgroups in Clinical Trial Design and Analysis
	1 Issues Related to Subgroup Analyses and Use of Intensive Stratification
		1.1 Introduction
		1.2 Issues in Interpreting Subgroup Data
			1.2.1 Confounding Due to Sampling Error
			1.2.2 Confounding Due to Repeated Hypothesis Testing
			1.2.3 Confounding Due to Lack of Statistical Power
			1.2.4 Confounding Due to Baseline Incomparability
		1.3 Intensive Stratification in Subgroup Analysis
		1.4 Statistical Simulations
			1.4.1 Setting
			1.4.2 Results
		1.5 Conclusion
		References
	2 Biomarker-Targeted Confirmatory Trials
		2.1 Introduction
		2.2 Inference Errors in an All-Comer Design
		2.3 Multiple-Step Inferences for the Overall Population and Marker Subgroups
		2.4 Simultaneous Inferences for the Overall Population and Marker-Positive Subgroup
		2.5 Extension to Sequential Analysis
		2.6 Discussion
		References
	3 Data-Driven and Confirmatory Subgroup Analysisin Clinical Trials
		3.1 Introduction
		3.2 Overview of Regulatory Guidance for Subgroup Analysis
			3.2.1 International Framework: ICH Guidelines
			3.2.2 Multi-Regional Clinical Trials: Case Study
			3.2.3 FDA Regulations
			3.2.4 EMA Regulations
			3.2.5 China National Medical Product Administration (NMPA) and Japanese Pharmaceuticals and Medical Devices Agency (PMDA) Regulations
		3.3 Data-Driven Subgroup Analysis
			3.3.1 Key Principles of Data-Driven Subgroup Analysis
			3.3.2 Types of Data-Driven Subgroup Analyses
			3.3.3 ``Guideline Driven\'\' Versus Data-Driven Subgroup Analysis
			3.3.4 Case Study
			3.3.5 Typology of Data-Driven Subgroup Analysis Methods
				3.3.5.1 Global Outcome Modeling
				3.3.5.2 Global Treatment Effect Modeling
				3.3.5.3 Modeling Individual Treatment Regimes
				3.3.5.4 Local Modeling (Direct Subgroup Search)
				3.3.5.5 Summary of Key Features of Data-Driven Methods
		3.4 Confirmatory Subgroup Analysis
			3.4.1 Multi-Population Trials
			3.4.2 Multiplicity Issues in Multi-Population Trials
				3.4.2.1 Traditional Multiplicity Problems
				3.4.2.2 Advanced Multiplicity Problems
				3.4.2.3 Clinical Scenario Evaluation Approach
			3.4.3 Decision-Making Framework in Multi-Population Trials
			3.4.4 Adaptive Designs in Multi-Population Trials
				3.4.4.1 Adaptive Trials with Pre-planned Subpopulations
				3.4.4.2 Adaptive Trials with Data-Driven Subpopulations
		3.5 Discussion
		References
	4 Considerations on Subgroup Analysis in Design and Analysis of Multi-Regional Clinical Trials
		4.1 Introduction
		4.2 Trial Design Considerations
			4.2.1 Region Definition
			4.2.2 Regional Requirements
			4.2.3 Randomization
		4.3 Models and the Overall Sample Sizes
		4.4 Considerations for Regional Subgroup Analysis
		4.5 Trial Conduct
		4.6 Interpretation of Results
		4.7 The Regular Subgroup Analyses
		4.8 Discussion
		References
Part II Subgroup Identification and Personalized Medicine
	5 Practical Subgroup Identification Strategies in Late-Stage Clinical Trials
		5.1 Introduction
		5.2 Case Study
		5.3 SIDES-Based Subgroup Identification Methods
			5.3.1 Subgroup Generation Algorithm
			5.3.2 Subgroup and Biomarker Selection Tools
			5.3.3 Subgroup Interpretation Tools
		5.4 Practical Considerations in Subgroup Identification
			5.4.1 Candidate Biomarkers
			5.4.2 Primary Analysis Model
		5.5 Subgroup Search Strategies in the BPH Trial
			5.5.1 Less Formal Subgroup Search Strategy
			5.5.2 More Formal Subgroup Search Strategy
		5.6 Simulation Study
		5.7 Discussion
		References
	6 The GUIDE Approach to Subgroup Identification
		6.1 Introduction
		6.2 Univariate Uncensored Response
			6.2.1 Node Models
			6.2.2 Split Variable Selection
			6.2.3 Split Set Selection
				6.2.3.1 Ordinal Variable
				6.2.3.2 Categorical Variable
		6.3 Bootstrap Confidence Intervals
		6.4 Multivariate Uncensored Responses
		6.5 Time-to-Event Response
		6.6 Concluding Remarks
		References
	7 A Novel Method of Subgroup Identification by Combining Virtual Twins with GUIDE (VG) for Development of Precision Medicines
		7.1 Introduction
		7.2 Methods
			7.2.1 Step I
			7.2.2 Step II
		7.3 Simulations
			7.3.1 Set-up
			7.3.2 Results
		7.4 Case Study
			7.4.1 Type I Error Control
			7.4.2 Bootstrap
			7.4.3 Application
		7.5 Discussion
		References
	8 Subgroup Identification for Tailored Therapies: Methods and Consistent Evaluation
		8.1 Background
		8.2 A Resampling-Based Ensemble Tree Method to Identify Patient Subgroups with Enhanced Treatment Effect
		8.3 Consistent Assessment of Biomarker and Subgroup Identification Methods
			8.3.1 Data Generation
			8.3.2 Performance Measurement
		8.4 Simulation Study
		8.5 Concluding Remarks
		References
	9 A New Paradigm for Subset Analysis in RandomizedClinical Trials
		9.1 Introduction
		9.2 Methods
			9.2.1 Predictive Classifiers
			9.2.2 De-biasing the Re-substitution Estimates
			9.2.3 Pre-validated Estimates of Treatment Effect
			9.2.4 Testing Treatment Effects in Subsets  S2 ,  S1  and  S0
			9.2.5 PPV and NPV of the Predictive Classifier
			9.2.6 Calibration of Pre-Validated Treatment Effects
		9.3 Discussion
		References
	10 Logical Inference on Treatment Efficacy When Subgroups Exist
		10.1 Introduction
		10.2 Fundamental Statistical Considerations When Subgroups Exist
			10.2.1 Treatment Efficacy Measures
				10.2.1.1 Which Group(s) Need to Be Assessed?
				10.2.1.2 Logic-Respecting Efficacy Measures
				10.2.1.3 Prognostic or Predictive?
			10.2.2 Inference on Mixture Populations
				10.2.2.1 Marginal Means
				10.2.2.2 LSmeans
		10.3 Subgroup Mixable Inference Procedure
			10.3.1 The General SME Principle
			10.3.2 Simultaneous Confidence Intervals
			10.3.3 Application of SME on Binary Outcomes
				10.3.3.1 Theoretical Derivations
				10.3.3.2 A Real Example
			10.3.4 Application of SME on Time-to-Event Outcomes
				10.3.4.1 Theoretical Derivations
				10.3.4.2 A Real Example
		10.4 Discussion
			10.4.1 Additional Issues or Challenges
			10.4.2 Moving Forward
		References
	11 Subgroup Analysis with Partial Linear Regression Model
		11.1 Introduction
		11.2 The Method
			11.2.1 The Semiparametric Model Specification
			11.2.2 Estimation of Model Parameters
				11.2.2.1 Computation of g(r+1)
			11.2.3 Asymptotic Results of the Estimates
		11.3 Testing the Null Hypothesis and the Classification Rules
			11.3.1 Test the Null Hypothesis
			11.3.2 The Classification Rule
		11.4 Simulation Study and Application
			11.4.1 Simulation Study
			11.4.2 Application to Real Data Problem
		11.5 Conclusion
		References
	12 Exploratory Subgroup Identification for Biopharmaceutical Development
		12.1 Introduction
		12.2 Single Biomarker Signature
			12.2.1 Prognostic Biomarker Signature Analysis
			12.2.2 Predictive Biomarker Signature Analysis
			12.2.3 A Framework for Robust Cutoff Derivation
		12.3 Complex Biomarker Signature
			12.3.1 Scoring-Based Methods
			12.3.2 Rule-Based Methods
				12.3.2.1 Sequential BATTing
				12.3.2.2 AIM-Rule
		12.4 Model Evaluation: Nested Cross-Validation
		12.5 Optimizing Long-Term Treatment Strategy: An Example of Subgroup Identification Leading to Label Inclusion
			12.5.1 Motivation and Background
			12.5.2 Method
			12.5.3 Result
		12.6 Discussion
		References
	13 Statistical Learning Methods for Optimizing Dynamic Treatment Regimes in Subgroup Identification
		13.1 Introduction
		13.2 Conceptual Framework
			13.2.1 Q-learning
			13.2.2 Outcome-Weighted Learning
				13.2.2.1 Outcome-Weighted Learning Without Augmentation
				13.2.2.2 Augmented Outcome-Weighted Learning (AOL)
				13.2.2.3 Surrogate Loss Functions in Outcome-Weighted Learning
			13.2.3 Evaluation of DTRs
		13.3 Estimation and Algorithm
			13.3.1 Q-learning
			13.3.2 Outcome-Weighted Learning
				13.3.2.1 Under SVM Hinge Loss
				13.3.2.2 Under SVM Ramp Loss
				13.3.2.3 Under Binomial Deviance Loss
				13.3.2.4 Under L2 Loss
				13.3.2.5 Tuning Parameters
			13.3.3 Handling Observational Study Data
			13.3.4 DTR Evaluation and Future Use
				13.3.4.1 Empirical Value Function and Benefit Function
				13.3.4.2 Apply the Learned DTR to an Independent Sample
		13.4 Software and Illustrations
			13.4.1 DTR Estimation
				13.4.1.1 Q-learning
				13.4.1.2 Outcome-Weighted Learning
			13.4.2 Apply the Estimated DTR to an Independent Sample
			13.4.3 Other Details
		13.5 Simulations and Real Data Implementation
			13.5.1 Simulations
			13.5.2 Illustration of the Real Data Implementation
		13.6 Summary
		References
Part III General Issues About Subgroup Analysis, Including Regulatory Considerations
	14 Subgroups in Design and Analysis of Clinical Trials, General Considerations
		14.1 General Issues in Subgroup Analysis as Part of the Overall Evaluation of a Clinical Trial
			14.1.1 Successful Trial for the Overall Study Population
			14.1.2 Failed Trial for the Intended Overall Patient Population
		14.2 Trial Design Considerations to Establish Treatment Efficacy in Specific Subgroup of Patients
		14.3 Bayesian Subgroup Analysis
		14.4 Summary
		References
	15 Subgroup Analysis: A View from Industry
		15.1 Introduction
		15.2 Defining a Subgroup Effect
		15.3 Multiplicity in Subgroup Analysis
		15.4 Statistical Methods
			15.4.1 Separate Analysis by Subgroup
			15.4.2 Interaction Tests
			15.4.3 Stepwise Regression
			15.4.4 Fractional Polynomial Modelling Approaches with Continuous Covariates
			15.4.5 Splines
			15.4.6 Shrinkage Methods
			15.4.7 Bayesian Dynamic Borrowing
			15.4.8 Partitioning Methods
		15.5 Discussion
		References
	16 Subgroup Analysis from Bayesian Perspectives
		16.1 Introduction
		16.2 Bayesian Subgroup Analysis Methods
			16.2.1 Tree-Based Bayesian Subgroup Analysis Methods
			16.2.2 ANOVA-Based Bayesian Subgroup Analysis Methods
			16.2.3 Other Types of Bayesian Subgroup Analysis Methods
		16.3 Simulation Studies
		16.4 A Real Data Example
		16.5 Discussion
		References
	17 Power of Statistical Tests for Subgroup Analysisin Meta-Analysis
		17.1 Overview of Subgroup Analysis in Meta-Analysis
			17.1.1 Potential Role of Power in Subgroup Analysis in Meta-Analysis
		17.2 Power Computations in Meta-Analysis
			17.2.1 Power for Subgroup Analyses in Meta-Analysis: Test of Between-Group Homogeneity in Fixed Effects Models
			17.2.2 Choosing Parameters for the Power of QB in Fixed Effects Models
			17.2.3 Example: Power for a Fixed Effects Analysis with the Standardized Mean Difference
			17.2.4 Example: Power for a Fixed Effects Analysis for the Log-Odds Ratio
			17.2.5 Power for Subgroup Analyses in Meta-Analysis: Test of Between-Group Homogeneity in Random Effects Models
			17.2.6 Choosing Parameters for the Power of QB in Random Effects Models
			17.2.7 Example: Power for a Random Effects Analysis with the Standardized Mean Difference
			17.2.8 Example: Power for a Random Effects Analysis for the Log-Odds Ratio
			17.2.9 Example: Unbalanced Number of Studies Within Subgroups
			17.2.10 Power for Other Tests of Moderators in Meta-Analysis
		17.3 Summary of Power for Subgroup Analysis in Meta-Analysis
		Appendix1 R Function pchisq to Compute Power
		Appendix2 R Code for Fig. 17.1, Power for Subgroup Differences with the Standardized Mean Difference
		Appendix3 R Code for Fig. 17.2, Power for Subgroup Differences with Log-Odds Ratio and Varying Degrees of Heterogeneity
		Appendix4 R Code for Fig. 17.3, Power for the Log-Odds Ratio with Varying Numbers of Studies Within Groups
		References
	18 Heterogeneity and Subgroup Analysis in Network Meta-Analysis
		18.1 Background
		18.2 Criteria for Valid Network Meta-Analysis
		18.3 Standard Network Meta-Analysis Model
		18.4 Specific Challenges with Subpopulations
		18.5 Shrinkage Estimation
		18.6 Network Meta-Regression
		18.7 Hierarchical Approach to Network Meta-Regression
		18.8 Conclusion
		References
Index