Applied Meta-Analysis with R and Stata

Cover
Half Title
Series Page
Title Page
Copyright Page
Dedication
Table of Contents
Preface for the Second Edition
Preface for the First Edition
Authors
List of Figures
List of Tables
1 Introduction to R and Stata for Meta-Analysis
	1.1 Introduction to R for Meta-Analysis
		1.1.1 What is R?
		1.1.2 Steps on Installing R and Updating R Packages
			1.1.2.1 First Step: Install R Base System
			1.1.2.2 Second Step: Installing and Updating R Packages
			1.1.2.3 Steps to Get Help and Documentation
		1.1.3 Database Management and Data Manipulations
			1.1.3.1 Data Management with RMySQL
			1.1.3.2 Data Management with Microsoft Excel and R Package gdata
			1.1.3.3 Data Management with Microsoft Excel and R Package xlsx
			1.1.3.4 Data Management with Microsoft Excel and R Package readxl
			1.1.3.5 Other Methods to Read Data into R
			1.1.3.6 R Package foreign
	1.2 A Simple Simulation on Multicenter Studies for Meta-Analysis
		1.2.1 Data Simulation
			1.2.1.1 R Functions
			1.2.1.2 Data Generation and Manipulation
			1.2.1.3 Basic R Graphics
		1.2.2 Data Analysis
			1.2.2.1 Data Analysis from Each Center
			1.2.2.2 Data Analysis with Pooled Data from Five Centers
			1.2.2.3 A Brief Introduction to Meta-Analysis
	1.3 Introduction of Stata for Meta-Analysis
	1.4 Summary and Recommendations for Further Reading about Using R
2 Research Protocol for Meta-Analyses
	2.1 Introduction
	2.2 Defining the Research Objective
	2.3 Criteria for Identifying Studies to Include in the Meta-Analysis
		2.3.1 Clarifying the Disease under Study (What Is Meant by Mild-to-Moderate?)
		2.3.2 The Effectiveness Measure or Outcome
		2.3.3 The Type of Control Group
		2.3.4 Study Characteristics
		2.3.5 Type of Patient
		2.3.6 Length of Study
	2.4 Searching for and Collecting the Studies
	2.5 Data Abstraction and Extraction
	2.6 Meta-Analysis Methods
	2.7 Results
	2.8 Summary and Discussion
3 Fixed Effects and Random Effects in Meta-Analysis
	3.1 Two Data Sets from Clinical Studies
		3.1.1 Data for Cochrane Collaboration Logo: Binary Data
		3.1.2 Clinical Studies on Amlodipine: Continuous Data
	3.2 Fixed-Effects and Random-Effects Models in Meta-Analysis
		3.2.1 Hypotheses and Effect Size
		3.2.2 Fixed-Effects Meta-Analysis Model: The Weighted-Average
			3.2.2.1 Fixed-Effects Model
			3.2.2.2 The Weighting Schemes
		3.2.3 Random-Effects Meta-Analysis Model: DerSimonian-Laird
			3.2.3.1 Random-Effects Model
			3.2.3.2 Derivation of DerSimonian-Laird Estimator of r[sup(2)]
		3.2.4 Publication Bias
	3.3 Meta-Analysis for Data from Cochrane Collaboration Logo
		3.3.1 The Data
		3.3.2 Fitting the Fixed-Effects Model
		3.3.3 Fitting the Random-Effects Model
	3.4 Meta-Analysis of Amlodipine Trial Data
		3.4.1 The Data
		3.4.2 Meta-Analysis with meta Package
			3.4.2.1 Fit the Fixed-Effects Model
			3.4.2.2 Fit the Random-Effects Model
		3.4.3 Meta-Analysis with the metafor Package
			3.4.3.1 Calculate the Effect Size
			3.4.3.2 Fit the Fixed-Effects Model
			3.4.3.3 Fit the Random-Effects Model
	3.5 Which Model Should We Use? Fixed Effects or Random Effects?
		3.5.1 Fixed-Effects
		3.5.2 Random-Effects
		3.5.3 Performing Both a Fixed-Effects and a Random-Effects Meta-Analysis
	3.6 Summary and Conclusions
	Appendix : Stata Programs for Fixed-Effects and Random-Effects in Meta-Analysis by Manyun Liu
4 Meta-Analysis with Binary Data
	4.1 Data from Real Life Studies
		4.1.1 Statin Clinical Trials
		4.1.2 Five Studies on Lamotrigine for Treatment of Bipolar Depression
	4.2 Meta-Analysis Methods
		4.2.1 Analysis with RR
			4.2.1.1 Definition
			4.2.1.2 Statistical Significance
			4.2.1.3 The RR Meta-Analysis: Step-by-Step
			4.2.1.4 RR Meta-Analysis with R package meta
			4.2.1.5 RR Meta-Analysis with R package metafor
		4.2.2 Analysis with Risk Difference
			4.2.2.1 Definition
			4.2.2.2 Meta-Analysis with Step-by-Step Implementation
			4.2.2.3 Meta-Analysis in R Package meta
			4.2.2.4 Meta-Analysis in R Package metafor
		4.2.3 Meta-Analysis with OR
			4.2.3.1 Data Structure
			4.2.3.2 OR: Woolf\'s Method
			4.2.3.3 Meta-Analysis with R Package meta
			4.2.3.4 Meta-Analysis with R Package metafor
		4.2.4 Meta-Analysis Using Mantel-Haenszel Method
			4.2.4.1 Details of the Mantel-Haenszel Method
			4.2.4.2 Step-by-Step R Implementation
			4.2.4.3 Meta-Analysis Using R Library Meta
		4.2.5 Peto\'s Meta-Analysis Method
			4.2.5.1 Peto\'s Odds Ratio
			4.2.5.2 Step-by-Step Implementation in R
			4.2.5.3 R Implementation in meta
	4.3 Meta-Analysis of Lamotrigine Studies
		4.3.1 Risk Ratio
		4.3.2 Risk Difference
		4.3.3 Odds Ratio
	4.4 Discussions
	Appendix: Stata Programs for Meta-Analysis with Binary Data by Manyun Liu
5 Meta-Analysis for Continuous Data
	5.1 Two Published Data Sets
		5.1.1 Impact of Intervention
		5.1.2 Tubeless vs Standard PCNL
	5.2 Methods for Continuous Data
		5.2.1 Estimate the MD △
		5.2.2 Estimate the SMD ó
	5.3 Meta-Analysis of the Impact of Intervention
		5.3.1 Load the Data Into R
		5.3.2 Meta-Analysis Using R Library meta
		5.3.3 Step-by-Step Implementation in R
		5.3.4 Meta-Analysis Using R Library metafor
	5.4 Meta-Analysis of Tubeless vs Standard PCNL
		5.4.1 Comparison of Operation Duration
		5.4.2 Comparison of Length of Hospital Stay
		5.4.3 Comparison of Postoperative Analgesic Requirement
		5.4.4 Comparison of Postoperative Hematocrit Change
		5.4.5 Conclusions and Discussion
	5.5 Discussion
	Appendix: Stata Programs for Meta-Analysis for Continuous Data by Manyun Liu
6 Heterogeneity in Meta-Analysis
	6.1 Heterogeneity Quantity Q and the Test of heterogeneity
	6.2 Quantifying Heterogeneity
		6.2.1 The τ[sup(2)] Index
		6.2.2 The H Index
		6.2.3 The I[sup(2)] Index
	6.3 Illustration with Cochrane Collaboration Logo Data – Binomial Data
		6.3.1 Cochrane Collaboration Logo Data
		6.3.2 Illustration Using R Package meta
		6.3.3 Implementation in R: Step by Step
		6.3.4 Meta-Analysis Using R Package metafor
	6.4 Illustration with PCNL Meta-Data – Continuous Data
		6.4.1 Tubeless vs Standard PCNL Data
		6.4.2 Implementation in R Library meta
		6.4.3 Implementation in R: Step by Step
		6.4.4 Implementation in R Library metafor
	6.5 Discussions
	Appendix: Stata Programs for Heterogeneity Assessment by Manyun Liu
7 Meta-Regression
	7.1 Data
		7.1.1 BCG Vaccine Data
		7.1.2 Ischemic Heart Disease
		7.1.3 ADHD for Children and Adolescents
	7.2 The Methods in Meta-Regression
	7.3 Meta-Analysis of BCG Data
		7.3.1 Random-Effects Meta-Analysis
		7.3.2 Meta-Regression Analysis
		7.3.3 Meta-Regression vs Weighted Regression
	7.4 Meta-Analysis of IHD Data
		7.4.1 IHD Data
		7.4.2 Random-Effects Meta-Analysis
		7.4.3 Meta-Regression Analysis
		7.4.4 Comparison of Different Fitting Methods
	7.5 Meta-Analysis of ADHD Data
		7.5.1 Data and Variables
		7.5.2 Meta-Analysis
		7.5.3 Meta-Regression Analysis
		7.5.4 Summary of ADHD Meta-Analysis
	7.6 Discussion
	Appendix: Stata Programs for Meta-Regression by Manyun Liu
8 Multivariate Meta-Analysis
	8.1 The Model and the Package of mvmeta
	8.2 Bivariate Meta-Data from Five Clinical Trials on Periodontal Disease
	8.3 Meta-Analysis with R Package mvmeta
		8.3.1 Fixed-Effects Meta-Analysis
		8.3.2 Random-Effects Meta-Analysis
		8.3.3 Meta-Regression
	8.4 Meta-Analysis with R Package metafor
		8.4.1 Rearrange the Data Format
		8.4.2 Fixed-Effects Meta-Analysis
		8.4.3 Random-Effects Meta-Analysis
		8.4.4 Meta-Regression
	8.5 Discussions
	Appendix: Stata Programs for Multivariate Meta-Analysis by Manyun Liu
9 Publication Bias in Meta-Analysis
	9.1 Introduction
	9.2 Reasons for Publication Bias in Systematic Review
	9.3 Dealing with Publication Bias
	9.4 Assessing the Potential of the Publication Bias
		9.4.1 R Codes
		9.4.2 Stata Codes
	9.5 Statistical Tests for Funnel-Plot Asymmetry
		9.5.1 Begg and Mazumdar\'s Rank Correlation Method
		9.5.2 Egger\'s Linear Regression Test
		9.5.3 Illustration with R and Stata System
			9.5.3.1 Using Data from Cochrane Collaboration Logo (Table 9.1)
			9.5.3.2 Using the Effect of Streptokinase After a Myocardial Infarction (strepto.dta) Data
	9.6 Other Issues of Publication Bias and Remedies
10 Strategies to Handle Missing Data in Meta-Analysis
	10.1 Introduction
	10.2 Strategies to Handle Missing Data in Meta-Analysis
		10.2.1 Deletion Methods
			10.2.1.1 List-Wise Deletion (Complete Case Analysis)
			10.2.1.2 Pair-Wise Deletion (Available Case Analysis)
		10.2.2 Single Imputation Methods
		10.2.3 Model-Based Methods For Missing Data
			10.2.3.1 Maximum Likelihood Method
			10.2.3.2 Multiple Imputations
	10.3 Sensitivity Analysis by Informative Missingness Approach
	10.4 Application
		10.4.1 Meta-Analysis of Studies With Missing Outcome
		10.4.2 Meta-Analysis of Studies With Missing Predictors
	10.5 Conclusion
11 Meta-Analysis for Evaluating Diagnostic Accuracy
	11.1 Introduction
	11.2 Medical Diagnostic Tests Accuracy Studies
	11.3 Meta-Analysis Pooling a Single Value of the Sensitivity or the Specificity
	11.4 Joint Meta-Analysis of Sensitivity and Specificity Resulting in a Summary Point
	11.5 Joint Meta-Analysis of Sensitivity and Specificity Resulting in Summary ROC Curve
	11.6 R Demonstration
	11.7 Stata Demonstration
	11.8 Other Statistical Packages for DTA
12 Network Meta-Analysis
	12.1 Concepts and Challenges of Network Meta-Analysis
	12.2 Data Sets
		12.2.1 Diabetes Clinical Trials
		12.2.2 Parkinson\'s Clinical Trials
	12.3 Frequentist Network Meta-Analysis
		12.3.1 Fixed-Effects Model
			12.3.1.1 Multiarm Studies
			12.3.1.2 I-Squared for Network Meta-Analysis
		12.3.2 Random-Effects Model
	12.4 Network Meta-Analysis of Diabetes Clinical Trial Data
		12.4.1 Network Meta-Analysis
		12.4.2 The Treatment Ranking
		12.4.3 Graphical Display of the Network Model
		12.4.4 Forest Plots
	12.5 The Net Heat Plot
	12.6 Bayesian Network Meta-Analysis
		12.6.1 Introduction to Bayesian Inference
			12.6.1.1 Baye\'s Theorem
			12.6.1.2 Prediction
			12.6.1.3 Bayesian Computation with Simulation
		12.6.2 Bayesian Meta-Analysis Models
			12.6.2.1 Fixed-Effects Models
			12.6.2.2 Random-Effects Models
		12.6.3 Bayesian Network Meta-Analysis Model
		12.6.4 Multiarm Trials
	12.7 Bayesian Network Meta-Analysis of Parkinson\'s Clinical Trial Data in R
		12.7.1 Data Preparation and Visualization
		12.7.2 Generate the Network Meta-Analysis Model
		12.7.3 Specify the Priors
		12.7.4 Markov Chain Monte Carlo Simulation
		12.7.5 Assessing the Convergence
		12.7.6 Assessing Inconsistency: The Nodesplit Method
		12.7.7 Summarize the Network Meta-Analysis Results
	12.8 Network Meta-Analysis of Parkinson\'s Clinical Trial Data in Stata
13 Meta-Analysis for Rare Events
	13.1 The Rosiglitazone Meta-Analysis
	13.2 Step-by-Step Data Analysis in R
		13.2.1 Load the Data
		13.2.2 Data Analysis for MI
		13.2.3 Data Analysis for Cardiovascular Death (Death)
	13.3 Discussion
14 Meta-Analyses with Individual Patient-Level Data versus Summary Statistics
	14.1 IPD with Five Studies of Lamotrigine
	14.2 Treatment Comparison for Changes in HAMD
		14.2.1 Meta-Analysis with IPD
			14.2.1.1 IPD Analysis by Each Study
			14.2.1.2 IPD Analysis with Pooled Data
			14.2.1.3 IPD Analysis Incorporating Covariates
			14.2.1.4 IPD Analysis with Linear Mixed-Effects Model
			14.2.1.5 Summary of IPD Analysis
		14.2.2 Meta-Analysis with SS
			14.2.2.1 Generate the SS
			14.2.2.2 Calculate the Effect Size: Mean Difference
			14.2.2.3 Meta-Analysis with Fixed-Effects Model
			14.2.2.4 Meta-Analysis with Random-Effects Model
	14.3 Treatment Comparison for Changes in MADRS
		14.3.1 Meta-Analysis with IPD
			14.3.1.1 IPD Analysis for Each Study
			14.3.1.2 IPD Analysis for All Four Studies
			14.3.1.3 IPD Analysis with Covariates
		14.3.2 Meta-Analysis with SS
			14.3.2.1 Generate SS
			14.3.2.2 Calculate ES: MD
			14.3.2.3 Fixed-Effects and Random-Effects Meta-Analyses
	14.4 Summary of Lamotrigine Analysis
	14.5 Simulation Study on Continuous Outcomes
		14.5.1 Simulation Data Generator
		14.5.2 Simulation Data Estimator
		14.5.3 Simulation
	14.6 Discussions
15 Other R/Stata Packages for Meta-Analysis
	15.1 R Packages of meta, rmeta, and metafor
	15.2 Combining p-Values in Meta-Analysis
	15.3 Combining Correlation Coefficients in Meta-Analysis
		15.3.1 R Package metacor
		15.3.2 Data on Land-Use Intensity
		15.3.3 Meta-Analysis Using metacor Package
		15.3.4 Meta-Analysis Using meta Package
		15.3.5 Further Discussions on Meta-Analysis for Correlations
	15.4 Other R Packages for Meta-Analysis
	15.5 Other Stata Methods for Meta-Analysis
Bibliography
Index