Exploratory Subgroup Analyses in Clinical Research

This essential guide on subgroup analyses in the emerging area of personalized medicine covers the issues of subgroup analyses from a practical and a theoretical/methodological point of view. The practical part introduces the issues using examples from the literature where subgroup analyses led to unexpected or difficult-to-interpret results, which have been interpreted differently by different stakeholders. On the technical side, the book addresses selection and selection bias variance reduction by borrowing information from the full population in estimating a subgroup effect. To this end, subgroup analysis will be linked to statistical modelling, and subgroup selection to model selection. This connection makes the techniques developed for model selection applicable to subgroup analysis. Beginning with a history of subgroup analysis, Exploratory Subgroup Analyses in Clinical Research offers chapters that cover: objectives and current practice of subgroup analyses; pitfalls of subgroup analyses; subgroup analysis and modeling; hierarchical models in subgroup analysis; and selection bias in regression. It also looks at the predicted individual treatment effect and offers an outlook of the topic in its final chapter. Focuses on the statistical aspects of subgroup analysis Filled with classroom and conference-workshop tested material Written by a leading expert in the field of subgroup analysis Complemented with a companion website featuring downloadable datasets and examples for teaching use Exploratory Subgroup Analyses in Clinical Research is an ideal book for medical statisticians and biostatisticians and will greatly benefit physicians and researchers interested in personalized medicine.

This essential guide on subgroup analyses in the emerging area of personalized medicine covers the issues of subgroup analyses from a practical and a theoretical/methodological point of view. The practical part introduces the issues using examples from the literature where subgroup analyses led to unexpected or difficult-to-interpret results, which have been interpreted differently by different stakeholders. On the technical side, the book addresses selection and selection bias variance reduction by borrowing information from the full population in estimating a subgroup effect. To this end, subgroup analysis will be linked to statistical modelling, and subgroup selection to model selection. This connection makes the techniques developed for model selection applicable to subgroup analysis.Beginning with a history of subgroup analysis, Exploratory Subgroup Analyses in Clinical Research offers chapters that cover: objectives and current practice of subgroup analyses; pitfalls of subgroup analyses; subgroup analysis and modeling; hierarchical models in subgroup analysis; and selection bias in regression. It also looks at the predicted individual treatment effect and offers an outlook of the topic in its final chapter.* Focuses on the statistical aspects of subgroup analysis* Filled with classroom and conference-workshop tested material* Written by a leading expert in the field of subgroup analysis* Complemented with a companion website featuring downloadable datasets and examples for teaching useExploratory Subgroup Analyses in Clinical Research is an ideal book for medical statisticians and biostatisticians and will greatly benefit physicians and researchers interested in personalized medicine.

Preface xiAcknowledgments xiiiAcronyms xvAbout the Companion Website xixIntroduction xxi1 Some History of Subgroup Analysis 11.1 Introduction 11.2 Questionable Subgroup Analyses 51.2.1 Star Signs May Matter 51.2.2 Unjustified Under-treatment 71.2.3 Misinterpretation of Center Effects 81.2.4 The End of a Career 101.3 Encouraging Subgroup Analyses 121.3.1 Higher Efficacy 121.3.2 Harm Prevention 131.3.3 Avoiding Unnecessary Treatment 161.4 Subgroups and Drug Approvals 181.4.1 A Convincing Subgroup 181.4.2 Inconsistencies Across Regions 191.4.3 Detecting Non-responders 221.4.4 In Search for Benefit 261.5 Concluding Remarks 292 Objectives and Current Practice of Subgroup Analyses 312.1 Introduction 312.2 Objectives of Subgroup Analyses 322.3 Definitions Around Subgroups 342.4 Confounding 372.5 Two Types of Subgroup Analyses 392.6 Reporting of Subgroups 432.7 Concluding Remarks 453 Pitfalls of Subgroup Analyses 473.1 Introduction 473.2 Extreme Effect Estimates 483.3 Selection Bias 513.4 Reversal of Effects 533.5 Regression to the Mean 573.6 Simpson's Paradox 603.7 Post-hoc Analyses 633.8 Concluding Remarks 654 Subgroup Analysis and Modeling 674.1 Introduction 674.2 Modeling and Prediction 694.3 Subgroups and Hierarchical Models 724.3.1 Stein's Discovery 724.3.2 The Normal-Normal Hierarchical Model 734.4 Subgroups and Regression Models 764.4.1 Subgroups Defined in Terms of Variables 764.4.2 The Predicted Individual Treatment Effect 794.4.3 Comparison of the Two Options 834.5 Variable Selection in Regression 854.5.1 Classical Variable Selection 864.5.2 Regularized Estimators 874.5.3 Variable Selection and Confounding 884.6 Concluding Remarks 895 Hierarchical Models in Subgroup Analysis 915.1 Introduction 915.2 A General Hierarchical Model 945.2.1 Robbins' Theorem and Tweedie's Formula 945.2.2 Mixture Priors 975.2.3 The False Discovery Rate 1005.3 Parameter Estimation 1015.3.1 Posterior Means and Variances 1015.3.2 Estimation Bias 1045.3.3 Selection Bias 1065.4 Case Studies 1115.4.1 The Toxoplasmosis Dataset 1115.4.2 The BCG Dataset 1135.4.3 The Prostate Cancer Dataset 1195.5 Concluding Remarks 1246 Selection Bias in Regression 1296.1 Introduction 1296.2 Correction for Selection Bias 1316.3 Variance Estimation 1366.4 A Case Study 1396.5 Concluding Remarks 1447 The Predicted Individual Treatment Effect 1477.1 Introduction 1487.2 Definition of the PITE 1497.3 Confidence Intervals of the PITE 1507.3.1 MLE for the Full Model 1517.3.2 MLE Under a Reduced Model 1517.3.3 Scheffé Confidence Bounds 1527.3.4 LASSO with Post-selection Intervals 1527.3.5 Randomized LASSO 1547.3.6 Simulation Study 1547.3.7 Extension to Other Endpoints 1577.4 Case Studies 1597.4.1 An Alzheimer Dataset 1607.4.2 The Prostate Cancer Study Again 1617.4.3 Renal Safety of Contrast Media 1657.5 Concluding Remarks 1738 Prediction models 1758.1 Introduction 1768.2 Prediction Error 1778.3 Model Selection or Averaging 1808.4 Prediction Error of the PITE 1828.5 A Case Study 1878.6 Concluding Remarks 1909 Outlook 193Bibliography 197Index 217