Applied Surrogate Endpoint Evaluation Methods with SAS and R - Applied Surrogate Endpoint Evaluation Methods with SAS and R -

Preface Introduction Overview of Surrogate Endpoint Evaluation Individual-Level versus Trial-Level Surrogacy Early Successes with Surrogates Early Failures with Surrogates Why Surrogate Endpoints Are Important Today The Need for Statistical Evaluation of Surrogates Surrogate Evaluation and Data Transparency Typical Uses of Surrogate Endpoints Earlier Clinical Endpoints Multiple Rating Scales More Sensitive Clinical Endpoints Biomarker-Based Surrogates Regulatory Use for Accelerated Approval Health Technology Assessment Structure of This Book Notation and Example Datasets Notation Example Datasets The Age-Related Macular Degeneration (ARMD) Trial Five Clinical Trials in Schizophrenia Advanced Ovarian Cancer: A Meta-Analysis of Four Clinical Trials Advanced Colorectal Cancer: A Meta-Analysis of 25 Trials Advanced Colorectal Cancer: A Meta-Analysis of 13 Trials Advanced Gastric Cancer: A Meta-Analysis of 20 Trials Advanced Prostate Cancer: A Meta-Analysis of Two Trials Acknowledgments for Use of Data The History of Surrogate Endpoint Evaluation: Single-Trial Methods Introduction Prentice’s Approach Definition Analysis of Case Studies: The Age-Related Macular Degeneration Trial An Appraisal of Prentice’s Approach The Proportion of Treatment Effect Explained Definition Analysis of Case Studies: The Age-Related Macular Degeneration Trial An Appraisal of the Proportion Explained The Relative Effect and Adjusted Association Definition Analysis of Case Studies: The Age-Related Macular Degeneration Trial An Appraisal of the Adjusted Association and the Relative Effect Issues with the Adjusted Association Issues with the Relative Effect Should the Single-Trial Methods Be Used in Practice?II Contemporary Surrogate Endpoint Evaluation Methods: Multiple-Trial Methods Two Continuous Outcomes Introduction The Meta-Analytic Framework Trial-Level Surrogacy Individual-Level Surrogacy Simplified Model-Fitting Strategies The Trial Dimension: Fixed- versus Random-Effects Models The Model Dimension: Full versus Reduced Models The Endpoint Dimension: Univariate versus Bivariate Models The Measurement Error Dimension: Weighted versus Unweighted Models General Considerations in the Multiple-Trial Setting The choice of trial-level units The coding of the treatment effect A “good” surrogate Prediction of Treatment Effect: Surrogate Threshold Effect (STE)Case Study: The Age-Related Macular Degeneration Trial Trial-level surrogacy Individual-level surrogacy Two Failure-Time Endpoints Introduction Theoretical Background Individual-level surrogacy Trial-level surrogacy Analysis of Case Studies Using SAS Copula-Based Models Marginal Models Using R Concluding Remarks A Categorical (Ordinal) and a Failure- Time Endpoint Introduction Theoretical Background Analysis of a Case Study Using SAS Four-Category Tumor Response Binary Tumor Response Using R Concluding Remarks Appendix A Continuous (Normally-Distributed) and a Failure-Time Endpoint Introduction Theoretical Background Analysis of a Case Study Using SAS Copula-Based Models Marginal Models Using R Concluding Remarks Appendix A Longitudinal (Normally Distributed) and a Failure-Time Endpoint Introduction Theoretical Background Analysis of a Case Study Using SAS Joint-Model-Based Analysis Marginal Models Using R Concluding Remarks Evaluation of Surrogate Endpoints from an Information-Theoretic Perspective Introduction An Information-Theoretic Unification Information-Theoretic Approach: Trial Level Plausibility of Finding a Valid Surrogate: Trial Level Estimating R21 Asymptotic Confidence Interval for Rh Information-Theoretic Approach: Individual-Level Surrogacy General Setting S and T Both Continuous Case Study Analysis: The Age-Related Macular Degeneration Trial S and T Longitudinal Asymptotic Confidence Interval for Ri Remarks S and T Time-to-Event Variables Estimating Rh ind Case Study Analysis: Four Ovarian Cancer Trials S and T: Binary-Binary or Continuous-Binary Case Study Analysis: Five Trials in Schizophrenia The binary-binary setting Two Categorical Endpoints Introduction S and T: Binary-Ordinal Information-Theoretic Approach Individual-Level Surrogacy: Binary-Ordinal Trial-Level Surrogacy: Binary-Ordinal Setting Confidence Intervals Computational Aspects Separation: Categorical Variables Separation: Binary Variables Separation: Ordinal Variables Impact of Separation on Surrogate Evaluation Solution to Separation Issues Separation: Final Considerations Surrogate Package: Binary-Ordinal Setting Case Study Analysis: Five Trials in Schizophrenia Summary: Binary-Ordinal Setting S and T: Ordinal-Binary or Ordinal-Ordinal Information-Theoretic Approach Individual-Level Surrogacy: Ordinal-Binary or Ordinal- Ordinal Settings Trial-Level Surrogacy: Ordinal-Binary or Ordinal-Ordinal Settings Computational Aspects Separation: Ordinal-Binary/Ordinal Setting Summary: Ordinal-Binary or Ordinal-Ordinal Setting Concluding Remarks III Software Tools SAS Software Introduction General Structure of the SAS Macros Available for the Analysis of Surrogate Endpoints Validation of Surrogacy Using a Joint Modeling Approach for Two Normally Distributed Endpoints The Full Fixed-Effects Model Model Formulation Sensitivity Analysis: Leave-One-Out Evaluation The SAS Macro %CONTCONTFULL Data Analysis and Output SAS Code for the First Step SAS Code for the Second Step The Reduced Fixed-Effects Model Model Formulation The SAS Macro %CONTCONTRED Data Analysis and Output SAS Code for the First Step The Full Mixed-Effects Model The SAS Macro %CONTRANFULL Data Analysis and Output SAS Code for the Full Mixed-Effects Model Reduced Mixed-Effects Model The SAS Macro %CONTRANRED Data Analysis and Output SAS Code for the Reduced Mixed-Effects Model Analysis for a Surrogacy Setting with Two Survival Endpoints A Two-Stage Approach (I)The SAS Macro %TWOSTAGECOX Data Analysis and Output SAS Code for the First-Stage Model A Two-Stage Approach (II)The SAS Macro %TWOSTAGEKM SAS Code for Trial-Specific KM Estimates (at a Given Time Point)A Joint Model for Survival Endpoints The SAS Macro '/.COPULA Validation Using Joint Modeling of a Time-to- Event and a Binary Endpoint Data Structure The SAS Macro %SURVBIN Data Analysis and Output The SAS Macro °/„SURVCAT A Continuous (Normally Distributed) and a Survival Endpoint Model Formulation Data Structure The SAS Macro °/„NORMSURV Data Analysis and Output Validation Using a Joint Model for Continuous and Binary Endpoints Data Structure The SAS Macro %NORMALBIN Data Analysis and Output SAS Code for the First-Stage Model Validation Using a Joint Model for Two Binary Endpoints Data Structure The SAS Macro %BINBIN Data Analysis and Output SAS Code for the First-Stage Model Validation Using the Information-Theory Approach Individual-Level Surrogacy Trial-Level Surrogacy Evaluation of Surrogate Endpoint for Two Continuous Endpoints The SAS Macro %NORMNORMINFO Data Analysis and Output Evaluation of Surrogacy for Survival and Binary Endpoints The SAS Macro %SURVBININFO Other Surrogacy Settings The R Package Surrogate Introduction Two Normally Distributed Endpoints The Meta-Analytic Framework Analyzing the Age-Related Macular Degeneration Dataset The Information-Theoretic Framework Analyzing the Age-Related Macular Degeneration Dataset Two Time-to-Event Endpoints Analyzing the Ovarian Cancer Dataset Two Binary Endpoints Analyzing the Data of Five Clinical Trials in Schizophrenia A Binary and a Normally Distributed Endpoint Analyzing the Data of Five Clinical Trials in Schizophrenia Estimation of Trial-Level Surrogacy When Only Trial-Level Data Are Available Analyzing Ten Hypothetical Trials Cloud Computing The Surrogate Shiny App Two Continuous Endpoints: The Reduced Fixed- Effects Model Two Time-to-Event Endpoints: A Two-Stage Approach Information-Theoretic Approach Individual- and Trial-Level Surrogacy Information-Theoretic Approach for Two Continuous Endpoints Information-Theoretic Approach for Two Binary Endpoints IV Additional Considerations and Further Topics Surrogate Endpoints in Rare Diseases Introduction Convergence Problems in Fitting Linear Mixed- Effects Models A Simulation Study Simulation scenarios Balance in cluster size Outcomes of interest Results Model Convergence Issues and Multiple Imputation A Simulation Study Case Studies The Age-Related Macular Degeneration Trial Five Clinical Trials in Schizophrenia A Formal Basis for the Two-Stage Approach High-Dimensional Biomarkers in Drug Discovery: The QSTAR Framework Introduction: From a Single Trial to a HighDimensional Setting The QSTAR Framework and Surrogacy Data The ROS1 Project The EGFR Project Graphical Interpretation (I): The Association between a Gene and Bioactivity Accounting for the Effect of a Fingerprint Feature Modeling Approach The Joint Model Inference Graphical Interpretation (II): Adjusted Association and Conditional Independence Analysis of the EGFR and the ROS1 Projects Application to the EGFR Pro ject Application to the ROS1 Project The R Package IntegratedJM Identification of Biomarkers Analysis of One Gene Using the gls Function The IntegratedJM Shiny App Concluding Remarks Evaluation of Magnetic Resonance Imaging as a Biomarker in Alzheimer’s Disease Introduction Alzheimer’s Disease Magnetic Resonance Imaging and Histology Parameters The AD Mouse Model for MRI and Histology Data Two Levels of Surrogacy Evaluation of MRI Parameters as a Biomarker for Histology Features A Joint Model for MRI and Histology Genotype-Specific Individual-Level Surrogacy Disease-Level Surrogacy the MRI Project Data: Examples of Region- Specific Models The Motor Cortex: GFAP Staining and MRI-AK The Caudate-Putamen: GFAP Staining and MRI-AK The Surrogacy Map of the Brain Implementation in SAS Data Structure Age-Specific Parameters for Histology in the Wildtype Model Implementation in R Common Parameter for Histology in the Wildtype Group Age-Specific Parameters for Histology in the Wild- type Group Concluding Remarks