Signal Detection for Medical Scientists: Likelihood Ratio Test-based Methodology (Chapman & Hall/CRC Biostatistics Series)

Signal Detection for Medical Scientists: Likelihood Ratio Based Test-Based Methodology presents the data mining techniques with focus on likelihood ratio test (LRT) based methods for signal detection. It emphasizes computational aspect of LRT methodology and is pertinent for first-time researchers and graduate students venturing into this interesting field. The book is written as a reference book for professionals in pharmaceutical industry, manufactures of medical devices, and regulatory agencies. The book deals with the signal detection in drug/device evaluation, which is important in the post-market evaluation of medical products, and in the pre-market signal detection during clinical trials for monitoring procedures. It should also appeal to academic researchers, and faculty members in mathematics, statistics, biostatistics, data science, pharmacology, engineering, epidemiology, and public health. Therefore, this book is well suited for both research and teaching. Key Features: Includes a balanced discussion of art of data structure, issues in signal detection, statistical methods and analytics, and implementation of the methods. Provides a comprehensive summary of the LRT methods for signal detection including the basic theory and extensions for varying datasets that may be large post-market data or pre-market clinical trial data. Contains details of scientific background, statistical methods, and associated algorithms that a reader can quickly master the materials and apply methods in the book on one's own problems This book presents the data mining techniques with focus on likelihood ratio test (LRT) based methods for signal detection. It emphasizes computational aspect of LRT methodology and is pertinent for first-time researchers and graduate students venturing into this interesting field. Cover 1 Half Title 2 Title Page 4 Copyright Page 5 Dedication 6 Contents 10 Preface 16 I. LRT Methodology 20 1. Introduction 22 1.1 Background of Data Mining for Safety Signals 22 1.2 Post-market Safety Databases 23 1.3 openFDA for FAERS Database 29 2. Data Mining Methods for Signal Detection 30 2.1 Review of Common Frequentist Methods 31 2.1.1 Reporting odds ratio (ROR) 32 2.1.2 Proportional reporting ratio 33 2.1.3 Information component 34 2.1.4 Chi-squared test 35 2.2 Review of Common Bayes and Empirical Bayes 35 2.2.1 Bayesian confidence propagation neural network (BCPNN) 17 36 2.2.2 Multi-item gamma poisson shrinker (MGPS) 37 2.2.3 New IC 39 2.2.4 Simplified Bayes 39 2.2.5 Poisson-DP method 40 2.3 Notes and Discussion 42 2.4 Appendix 46 2.4.1 Relationship between new IC and sB methods 46 2.4.2 Polya-urn representation of the Dirichlet process prior—Conditional distributions 47 3. Basic LRT Method 50 3.1 Method Development 50 3.1.1 Derivation of likelihood ratio test statistic 51 3.1.2 MLR incorporating covariate information 52 3.1.3 Hypothesis testing 52 3.2 Simulation 54 3.2.1 Data simulation 54 3.2.2 Performance characteristics 56 3.2.3 Simulation results 56 3.3 Applications to FAERS Data 58 3.3.1 Montelukast analysis 59 3.3.2 Heparin analysis 59 3.4 Notes and Discussion 61 3.5 Further Discussion Points 63 3.6 LRT Tool in openFDA 65 3.7 Tables and Figures 66 4. LRT Methods for Drug Classes 72 4.1 Signals for Drug Class 72 4.1.1 Drug signals in FAERS data 72 4.1.2 Signals of single drug-AE combination 73 4.1.3 Performance evaluation using simulation 74 4.2 Ext-LRT Signal Detection using Double Maximum 75 4.2.1 Ext-LRT statistic and statistical inference 75 4.2.2 Application of Ext-LRT method 77 4.3 LRT for Identification of Signals Using Weight Matrix 78 4.3.1 Modi ed LRT statistic and statistical inference 78 4.3.2 Applications in detection of signals—Collection of drugs 80 4.4 Notes and Discussion 82 4.5 Appendix: Tree-based Scan Statistic Method for AE Signals Evaluation 84 4.5.1 Summary of the method 84 4.5.2 Relationship to the LRT method 84 4.6 Tables and Figures 85 5. ZIP-LRT Method for Modeling Extra-zeros 92 5.1 ZIP-LRT Model 92 5.2 EM Algorithm for ZIP Model 93 5.3 Likelihood Ratio Test Statistic 96 5.4 Strati ed ZIP LRT 96 5.5 Hypothesis Testing 97 5.6 Simulation Study 99 5.6.1 Data simulation 99 5.6.2 Simulation results 100 5.7 Application to 2006-2011 FAERS Data 100 5.7.1 Estimated percentage of true zeros 100 5.7.2 Application to selected drugs 101 5.8 Notes and Discussion 102 5.9 Appendix: Proof of Theorem 5.5.1 104 5.10 Tables and Figures 106 II. Extensions 110 6. LRT Method for Active Safety Surveillance with Exposure Information 112 6.1 Medical Background Based on the Example Data 112 6.1.1 Data structure 113 6.1.2 De nitions of drug exposure 113 6.1.3 De ning multiple looks 114 6.2 Longitudinal LRT Method for Active Safety Surveil Surveillance 115 6.2.1 LongLRT for comparing multiple events using event-time 115 6.2.2 SeqLRT for comparing two drugs and one AE with single occurrence using person-time 116 6.2.3 LongLRT for comparing multiple drugs and one AE with recurrence using exposure-time 117 6.2.4 Assumption of independence 118 6.3 Statistical Inference with Multiple Looks 119 6.4 Applications 120 6.4.1 Safety signals (among multiple AEs) by drug 120 6.4.2 Safety signals for the rst occurrence of a composite AE and two drugs using SeqLRT 121 6.4.3 Safety signals for multiple occurrences of a composite AE and two drugs using LongLRT 121 6.4.4 Safety signals for a composite AE with recurrence from multiple drugs using LongLRT 122 6.5 Simulation Study for Longitudinal LRT Methods 122 6.5.1 Data simulation 122 6.5.2 Performance characteristics 123 6.5.3 Simulation results 124 6.6 Discussion 124 6.7 Appendix 127 6.7.1 De nition of the composite AE 127 6.7.2 Data structure for the I x J matrix 127 6.7.3 Independence of the parameters 128 6.8 Tables and Figures 129 7. LRT-based Methodologies for Analysis of Multiple Studies 140 7.1 Background and Motivation 140 7.2 Methods 141 7.2.1 Summary of basic LRT with and without exposure information 141 7.2.2 LRT analysis approaches for signal detection from multiple studies 143 7.3 Applications 144 7.3.1 Analysis of PPI data with two drugs and a composite AE 145 7.3.2 Analysis of lipiodol data with one drug and multiple AEs 147 7.3.3 Summary of the two examples 148 7.4 Simulation 148 7.4.1 Simulation setup 149 7.4.2 Results 150 7.5 Discussion 151 7.6 Tables and Figures 153 III. Additional Frameworks 158 8. LRT Methods in Medical Device Safety Evaluation 160 8.1 Background 160 8.2 LRT Methods for Device Data from MDR 161 8.2.1 MDR data for LVADs 161 8.2.2 Exploration of the device data 162 8.2.3 Remarks 164 8.3 Safety Evaluation in Treatment vs. Control Group for Medical Device 164 8.3.1 Data source 164 8.3.2 Statistical models 165 8.3.3 Conventional Z-test with P-value adjustment 165 8.3.4 Modified LRT 166 8.3.5 Results 167 8.3.6 Performance evaluation using simulation 168 8.3.7 Remarks 169 8.3.8 Appendix: LRT and tree-based scan statistic method for comparing treatment vs. control 171 8.3.8.1 Summary of tree-based scan method 171 8.3.8.2 Relationship between LRT method and tree-based scan statistic method 172 8.4 Spatial-Cluster Signal Detection in Medical Devices using LRT 173 8.4.1 Background 173 8.4.2 Spatial-LRT with exposure information 174 8.4.3 Medical device safety database considerations 175 8.4.4 Illustrations 176 8.4.5 Remarks 177 8.5 Weighted LRT Method to Device Data From Multiple Sites 178 8.5.1 Background 178 8.5.2 Data structure for data with multiple studies 179 8.5.3 Weighted LRT method 179 8.5.4 Results 180 8.5.5 Remarks 180 8.6 Tables and Figures 181 9. LRT Method for Multiple-Site Device Data with Continuous Outcomes 194 9.1 Background 194 9.2 Data Structure and Problem Formulation 195 9.2.1 Data structure 195 9.2.2 Problem formulation 196 9.3 Normal-LRT method 196 9.3.1 MLRs of parameters uij ; vij , and σ2 j 196 9.3.2 Test statistic MLLR 197 9.3.3 Permutation-based empirical distribution 198 9.4 Application 199 9.5 Simulation 200 9.5.1 Data simulation 200 9.5.2 Type-I error 200 9.5.3 Power, sensitivity, and false discovery rate 200 9.6 Notes and Discussion 201 9.7 Appendix 203 9.7.1 Data generation process for hypothetical case study 203 9.7.2 Simulation for data including regions without data 203 9.8 Tables and Figures 205 10. Use of LRT in Site Selection 210 10.1 Background 210 10.2 Data and P-value Matrix 211 10.3 Statistical Approaches for Site Ranking Generation from Pvalue Matrix 212 10.3.1 Fisher combination approach 212 10.3.2 LRT approach 213 10.4 Application to a Case Study 214 10.5 Comprehensive Simulation Study 215 10.5.1 Performance characteristics 215 10.5.2 Results 216 10.6 Notes and Discussion 218 10.7 Appendix 220 10.7.1 Statistical tests for individual variables 220 10.7.2 Simulation of P-values for site signals 222 10.7.3 Simulation study for correlation P-values 223 10.8 Tables and Figures 225 Bibliography 230 Subject Index 244 LRT,Method;,Bayesian,Methods;,Adverse,Effect;,Data,Mining;,Outlier,detection LRT Method,Bayesian Methods,Adverse Effect,Data Mining,Outlier detection "This book presents the data mining techniques with focus on likelihood ratio test (LRT) based methods for signal detection. It emphasizes computational aspect of LRT methodology and is pertinent for first-time researchers and graduate students venturing into this interesting field. The book is written as a reference book for professionals in pharmaceutical industry, manufactures of medical devices, and regulatory agencies. The book deals with the signal detection in drug/device evaluation, which is important in the post-market evaluation of medical products, and in the pre-market signal detection during clinical trials for monitoring procedures"-- Provided by publisher