Statistical Remedies for Medical Researchers (Springer Series in Pharmaceutical Statistics)

"This book illustrates numerous statistical practices that are commonly used by medical researchers, but which have severe flaws that may not be obvious. For each example, it provides one or more alternative statistical methods that avoid misleading or incorrect inferences being made. The technical level is kept to a minimum to make the book accessible to non-statisticians. At the same time, since many of the examples describe methods used routinely by medical statisticians with formal statistical training, the book appeals to a broad readership in the medical research community." -- Prové de l'editor Preface......Page 7 Contents......Page 9 1 Why Bother with Statistics?......Page 12 1.1 Some Unexpected Problems......Page 13 1.2 Expert Opinion......Page 14 1.3 The Innocent Bystander Effect......Page 16 1.4 Gambling and Medicine......Page 20 1.5 Testing Positive......Page 23 1.6 Bayes' Law and Hemophilia......Page 27 2 Frequentists and Bayesians......Page 30 2.1 Statistical Inference......Page 31 2.2 Frequentist Statistics......Page 34 2.3 Bayesian Statistics......Page 39 3 Knocking Down the Straw Man......Page 50 3.1 Designing Clinical Trials......Page 51 3.2 A Common Phase II Design......Page 53 3.3 A Common Misinterpretation......Page 56 3.4 Not Testing Hypotheses......Page 58 3.5 Random Standards......Page 61 3.6 A Fake Null Hypothesis......Page 62 3.7 Monitoring Toxicity and Response......Page 66 4 Science and Belief......Page 70 4.1 Theory Versus Practice......Page 71 4.2 Technology and Cherry-Picking......Page 74 4.3 Is a New Treatment Any Good?......Page 79 4.4 The Las Vegas Effect......Page 82 5 The Perils of P-Values......Page 86 5.1 Counting Cows......Page 87 5.2 A Sacred Ritual......Page 88 5.3 A Dataset with Four P-Values......Page 93 5.4 Bayes Factors......Page 95 5.5 Computing Sample Sizes......Page 97 5.6 Not-So-Equivalent Studies......Page 101 5.7 Acute Respiratory Distress Syndrome......Page 102 5.8 The Multiple Testing Problem......Page 106 5.9 Type S Error......Page 113 5.10 A Simple Bayesian Alternative to P-Values......Page 115 5.11 Survival Analysis Without P-Values......Page 118 5.12 The P-Value War......Page 120 6 Flipping Coins......Page 125 6.1 Farming and Medicine......Page 126 6.2 How Not to Compare Treatments......Page 127 6.3 Counterfactuals and Causality......Page 129 6.4 Why Randomize?......Page 134 6.5 Stratifying by Subgroups......Page 136 6.6 Inverse Probability-Weighted Survival Analysis......Page 139 6.7 Bias Correction by Matching......Page 143 6.8 A Bayesian Rationale for Randomization......Page 148 6.9 Outcome-Adaptive Randomization......Page 150 7 All Mixed Up......Page 159 7.1 The Billion Dollar Computation......Page 160 7.2 Accounting for Uncertainty and Bias......Page 164 7.3 Predicting Phase III Success......Page 165 7.4 A Paradoxical Clinical Trial......Page 168 8 Sex, Biomarkers, and Paradoxes......Page 173 8.1 A Paradox......Page 174 8.2 Batting Averages......Page 176 8.3 A Magic Biomarker......Page 177 8.4 Plotting Regression Data......Page 181 9 Crippling New Treatments......Page 192 9.1 Phase I Trials......Page 193 9.2 Choosing the Wrong Dose in Phase I......Page 197 9.3 Phase I–II Designs......Page 204 10 Just Plain Wrong......Page 210 10.1 Clinical Trial Design, Belief, and Ethics......Page 211 10.2 A Futile Futility Rule......Page 212 10.3 The Evaluability Game......Page 216 10.4 The Fox, the Farmer, and the Chickens......Page 220 10.5 Planned Confounding......Page 222 10.6 Select-and-Test Designs......Page 227 11 Getting Personal......Page 233 11.1 From Bench to Bedside......Page 234 11.2 Age Discrimination......Page 236 11.3 Comparing Treatments Precisely......Page 240 11.4 A Subgroup-Specific Phase II–III Design......Page 245 11.5 Precision Pharmacokinetic Dosing......Page 247 12 Multistage Treatment Regimes......Page 254 12.1 The Triangle of Death......Page 255 12.2 Observe, Act, Repeat......Page 262 12.3 SMART Designs......Page 265 12.4 Repeat or Switch Away......Page 271 12.5 A Semi-SMART Design......Page 277 BookmarkTitle:......Page 284 Index......Page 295 Preface 7 Contents 9 1 Why Bother with Statistics? 12 1.1 Some Unexpected Problems 13 1.2 Expert Opinion 14 1.3 The Innocent Bystander Effect 16 1.4 Gambling and Medicine 20 1.5 Testing Positive 23 1.6 Bayes' Law and Hemophilia 27 2 Frequentists and Bayesians 30 2.1 Statistical Inference 31 2.2 Frequentist Statistics 34 2.3 Bayesian Statistics 39 3 Knocking Down the Straw Man 50 3.1 Designing Clinical Trials 51 3.2 A Common Phase II Design 53 3.3 A Common Misinterpretation 56 3.4 Not Testing Hypotheses 58 3.5 Random Standards 61 3.6 A Fake Null Hypothesis 62 3.7 Monitoring Toxicity and Response 66 4 Science and Belief 70 4.1 Theory Versus Practice 71 4.2 Technology and Cherry-Picking 74 4.3 Is a New Treatment Any Good? 79 4.4 The Las Vegas Effect 82 5 The Perils of P-Values 86 5.1 Counting Cows 87 5.2 A Sacred Ritual 88 5.3 A Dataset with Four P-Values 93 5.4 Bayes Factors 95 5.5 Computing Sample Sizes 97 5.6 Not-So-Equivalent Studies 101 5.7 Acute Respiratory Distress Syndrome 102 5.8 The Multiple Testing Problem 106 5.9 Type S Error 113 5.10 A Simple Bayesian Alternative to P-Values 115 5.11 Survival Analysis Without P-Values 118 5.12 The P-Value War 120 6 Flipping Coins 125 6.1 Farming and Medicine 126 6.2 How Not to Compare Treatments 127 6.3 Counterfactuals and Causality 129 6.4 Why Randomize? 134 6.5 Stratifying by Subgroups 136 6.6 Inverse Probability-Weighted Survival Analysis 139 6.7 Bias Correction by Matching 143 6.8 A Bayesian Rationale for Randomization 148 6.9 Outcome-Adaptive Randomization 150 7 All Mixed Up 159 7.1 The Billion Dollar Computation 160 7.2 Accounting for Uncertainty and Bias 164 7.3 Predicting Phase III Success 165 7.4 A Paradoxical Clinical Trial 168 8 Sex, Biomarkers, and Paradoxes 173 8.1 A Paradox 174 8.2 Batting Averages 176 8.3 A Magic Biomarker 177 8.4 Plotting Regression Data 181 9 Crippling New Treatments 192 9.1 Phase I Trials 193 9.2 Choosing the Wrong Dose in Phase I 197 9.3 Phase I–II Designs 204 10 Just Plain Wrong 210 10.1 Clinical Trial Design, Belief, and Ethics 211 10.2 A Futile Futility Rule 212 10.3 The Evaluability Game 216 10.4 The Fox, the Farmer, and the Chickens 220 10.5 Planned Confounding 222 10.6 Select-and-Test Designs 227 11 Getting Personal 233 11.1 From Bench to Bedside 234 11.2 Age Discrimination 236 11.3 Comparing Treatments Precisely 240 11.4 A Subgroup-Specific Phase II–III Design 245 11.5 Precision Pharmacokinetic Dosing 247 12 Multistage Treatment Regimes 254 12.1 The Triangle of Death 255 12.2 Observe, Act, Repeat 262 12.3 SMART Designs 265 12.4 Repeat or Switch Away 271 12.5 A Semi-SMART Design 277 Appendix References 284 284 Index 295