Clinical Epidemiology

ix
Introduction 1
Clinical Questions and Clinical Epidemiology 2
Health Outcomes 2
The Scientific Basis for Clinical Medicine 3
Basic Principles 5
Variables 6
Numbers and Probability 6
Populations and Samples 6
Bias (Systematic Error) 6
Chance 10
The Effects of Bias and Chance
Are Cumulative 10
Internal and External Validity 11
Information and Decisions 12
Organization of This Book 12
Frequency 17
Are Words Suitable Substitutes for Numbers? 17
Prevalence and Incidence 18
Prevalence 18
Incidence 18
Prevalence and Incidence
in Relation to Time 19
Relationships Among Prevalence, Incidence,
and Duration of Disease 19
Some Other Rates 20
Studies of Prevalence and Incidence 20
Prevalence Studies 21
Incidence Studies 21
Cumulative Incidence 21
Incidence Density (Person-Years) 21
Basic Elements of Frequency Studies 22
What Is a Case? Defining
the Numerator 22
What Is the Population? Defining
the Denominator 24
Does the Study Sample Represent the
Population? 25
Distribution of Disease by Time, Place,
and Person 25
Time 25
Place 26
Person 26
Uses of Prevalence Studies 27
What Are Prevalence Studies Good for? 27
What Are Prevalence Studies Not Particularly
Good for? 28
Abnormality 31
Types of Data 32
Nominal Data 32
Ordinal Data 32
Interval Data 32
Performance of Measurements 33
Validity 33
Reliability 34
Range 35
Responsiveness 35
Interpretability 35
1
Contents
2
3 Contents
ariation 36
Variation Resulting from Measurement 36
Variation Resulting from
Biologic Differences 36
Total Variation 37
Effects of Variation 38
Distributions 39
Describing Distributions 39
Actual Distributions 39
The Normal Distribution 39
Criteria for Abnormality 41
Abnormal = Unusual 42
Abnormal = Biologic Dysfunction 43
Abnormal = Illness 45
Abnormal = Treating the Condition Leads to a
Better Clinical Outcome 47
Regression to the Mean 48
Diagnosis 53
Simplifying Data 53
he Accuracy of a Test Result 54
The Gold Standard 55
ensitivity and Specificity 55
Definitions 55
Use of Sensitive Tests 55
Use of Specific Tests 57
Trade-Offs Between Sensitivity
and Specificity 57
The Receiver Operator
Characteristic (ROC) Curve 58
tudies of Diagnostic Tests 59
Spectrum of Patients—the Study Population 60
Bias 61
Chance 61
Imperfect Gold Standards 62
redictive Value 64
Definitions 64
Determinants of Predictive Value 65
Estimating Prevalence (Pretest Probability) 66
Implications for Interpreting the Medical
Literature 68
Likelihood Ratios 68
Odds 68
Definitions 69
Use of Likelihood Ratios 69
Why Use Likelihood Ratios? 69
Calculating Likelihood Ratios 70
Multiple Tests 71
Parallel Testing 72
Clinical Prediction Rules 73
Serial Testing 74
Serial Likelihood Ratios 74
Assumption of Independence 74
Risk: Basic
Principles 78
Risk Measurement 79
Risk Factors 79
Recognizing Risk Factors 80
Long Latency 80
Immediate Versus Distant Causes 80
Common Exposure to Risk Factors 80
Low Incidence of Disease 81
Small Risk 81
Multiple Causes and Multiple Effects 81
Risk Factors May or May Not Be Causal 81
Risk Prediction Models 82
Combining Multiple Factors 82
Evaluating Risk Prediction Tools 83
Discrimination 83
Calibration 85
Validating Models 86
External Validation 86
Comparing Models 87
Assessing Models in Clinical Practice 87
Risk Stratification 87
Clinical Uses of Risk Factors, Prognostic Factors,
and Risk Prediction Tools 88
Risk Prediction and Pretest Probability for
Diagnostic Testing 88
Using Risk Factors to Choose Treatment 89
Risk Stratification for Screening Programs 89
Removing Risk Factors to Prevent Disease 89
4
5 Contents xi
Risk: Exposure to
Disease 92
Studies of Risk 92
When Experiments Are Not Possible or Ethical 92
Cohorts 93
Cohort Studies 93
Prospective and Historical Cohort Studies 94
Advantages and Disadvantages
of Cohort Studies 96
Ways to Express and Compare Risk 98
Absolute Risk 99
Attributable Risk 99
Relative Risk 99
Interpreting Attributable and Relative Risk 99
Population Risk 100
Taking Other Variables into Account 101
Extraneous Variables 101
Simple Descriptions of Risk 101
Confounding 102
Working Definition 102
Potential Confounders 102
Confirming Confounding 102
Control of Confounding 103
Randomization 103
Restriction 103
Matching 104
Stratification 104
Standardization 105
Multivariable Adjustment 105
Overall Strategy for Control of Confounding 106
Observational Studies and Cause 106
Effect Modification 106
Mendelian Randomization 107
Risk: From Disease to
Exposure 111
Case-Control Studies 112
Design of Case-Control Studies 114
The Source Population 114
Selecting Cases 114
Selecting Controls 114
Measuring Exposure 116
The Odds Ratio: An Estimate of
Relative Risk 118
Odds Ratio Calculation 119
Odds Ratio as an Indirect
Estimate of Relative Risk 119
Odds Ratio as a Direct Estimate
of Relative Risk 120
Controlling for
Extraneous Variables 120
Investigation of a
Disease Outbreak 121
Prognosis 126
Differences in Risk and Prognostic
Factors 126
The Patients Are Different 127
The Outcomes Are Different 127
The Rates Are Different 127
The Factors May be Different 127
Clinical Course and Natural History
of Disease 127
Elements of Prognostic Studies 127
Patient Sample 127
Zero Time 128
Follow-Up 129
Outcomes of Disease 129
Describing Prognosis 129
A Trade-Off: Simplicity Versus
More Information 129
Survival Analysis 130
Survival of a Cohort 130
Survival Curves 132
Interpreting Survival Curves 133
Identifying Prognostic Factors 133
Case Series 134
Clinical Prediction Rules 134
Bias in Cohort Studies 135
Sampling Bias 136
Migration Bias 136
6
7
8xii Contents
Measurement Bias 136
Bias from “Non-differential”
Misclassification 137
Bias from Missing Data 137
Bias, Perhaps, But Does It Matter? 137
ensitivity Analysis 137
Treatment 142
Ideas and Evidence 142
Ideas 142
Testing Ideas 143
tudies of Treatment Effects 144
Observational and Experimental Studies
of Treatment Effects 144
andomized Controlled Trials 144
Ethics 145
Sampling 145
Intervention 147
Comparison Groups 147
Allocating Treatment 148
Differences Arising After Randomization 149
Blinding 150
Assessment of Outcomes 150
Efficacy and Effectiveness 152
Intention-to-Treat and
Explanatory Trials 153
Superiority, Equivalence, and Noninferiority 153
ariations on Basic Randomized Trials 155
ailoring the Results of Trials to Individual
atients 156
Subgroups 156
Effectiveness in Individual Patients 156
N of 1 Trials 156
Alternatives to Randomized Controlled
rials 157
Limitations of Randomized Trials 157
Observational Studies
of Interventions 157
Clinical Databases 158
Randomized Versus
Observational Studies? 158
Phases of Clinical Trials 158
Prevention 162
Preventive Activities in Clinical
Settings 162
Types of Clinical Prevention 163
Levels of Prevention 163
Primary Prevention 163
Secondary Prevention 164
Tertiary Prevention 164
Confusion About Primary, Secondary, and
Tertiary Prevention 164
Scientific Approach to Clinical
Prevention 165
Burden of Suffering 165
Effectiveness of Treatment 166
Treatment in Primary Prevention 166
Treatment in Secondary Prevention 167
Treatment in Tertiary Prevention 168
Methodologic Issues in Evaluating Screening
Programs 169
Prevalence and Incidence Screens 169
Special Biases 169
Performance of
Screening Tests 172
High Sensitivity and Specificity 172
Detection and Incidence Methods
for Calculating Sensitivity 173
Low Positive Predictive Value 174
Simplicity and Low Cost 174
Safety 175
Acceptable to Patients and
Clinicians 175
Unintended Consequences
of Screening 175
Risk of False-Positive Result 176
Risk of Negative Labeling Effect 176
Risk of Overdiagnosis (Pseudodisease) in
Cancer Screening 177
Incidentalomas 178
Changes in Screening Tests and Treatments
Over Time 179
Weighing Benefits Against Harms of
Prevention 179
9
10 Contents xiii
Chance 185
Two Approaches to Chance 185
Hypothesis Testing 186
False-Positive and False-Negative Statistical
Results 186
Concluding That a Treatment Works 186
Dichotomous and Exact P Values 187
Statistical Significance and
Clinical Importance 187
Statistical Tests 188
Concluding That a Treatment
Does Not Work 189
How Many Study Patients Are Enough? 190
Statistical Power 190
Estimating Sample Size Requirements 190
Point Estimates and Confidence Intervals 193
Statistical Power After a Study
Is Completed 194
Detecting Rare Events 194
Multiple Comparisons 194
Subgroup Analysis 196
Multiple Outcomes 197
Noninferiority Studies 198
Multivariable Methods 198
Bayesian Reasoning 200
Cause 204
Basic Principles 205
Single Causes 205
Multiple Causes 205
Proximity of Cause to Effect 206
Indirect Evidence for Cause 208
Examining Individual Studies 208
Hierarchy of Research Designs 209
The Body of Evidence for
and Against Cause 209
Does Cause Precede Effect? 210
Strength of the Association 210
Dose–Response Relationships 210
Reversible Associations 211
Consistency 211
Biologic Plausibility 211
Specificity 212
Analogy 212
Aggregate Risk Studies 212
Modeling 214
Weighing the Evidence 216
Summarizing the
Evidence 219
Traditional Reviews 219
Systematic Reviews 220
Defining a Specific Question 220
Selecting Studies 221
Assessing Study Quality and Characteristics 223
Summarizing Results 225
Combining Studies in Meta-Analyses 226
Are the Studies Similar Enough to Justify
Combining? 226
How Are the Results Pooled? 227
Identifying Reasons for Heterogeneity 228
Additional Meta-Analysis Methods 229
Patient-Level Meta-Analysis 229
Network Meta-Analysis 230
Cumulative Meta-Analyses 230
Systematic Reviews of Observational and
Diagnostic Studies 231
Strengths and Weaknesses of Meta-Analyses 232
Knowledge
Management 236
Basic Principles 236
Do It Yourself or Delegate? 236
Which Medium? 237
Grading Information 237
Misleading Reports of Research Findings 237
Looking Up Answers to Clinical Questions 239
Solutions 239
Surveillance on New Developments 241
11
12
13
14xiv Contents
Journals 242
“Reading” Journals 243
Guiding Patients’ Quest for Health
Information 245
Putting Knowledge Management
into Practice 245
APPENDIX A: ANSWERS TO REVIEW
QUESTIONS 249
APPENDIX B: ADDITIONAL READINGS 262
INDEX 265