Biostatistics and Epidemiology (eBook)

PREFACE TO THE THIRD EDITION 7
ACKNOWLEDGMENTS 10
CONTENTS 11
THE SCIENTIFIC METHOD 17
1.1 The Logic of Scientific Reasoning 17
1.2 Variability of Phenomena Requires 22
1.3 Inductive Inference: Statistics as the Technology of the Scientific Method 23
1.4 Design of Studies 24
1.5 How to Quantify Variables 26
1.6 The Null Hypothesis 27
1.7 Why Do We Test the Null Hypothesis? 28
1.8 Types of Errors 30
1.9 Significance Level and Types of Error 31
1.10 Consequences of Type I and Type II Errors 32
A LITTLE BIT OF PROBABILITY 34
2.1 What Is Probability? 34
2.2 Combining Probabilities 35
2.3 Conditional Probability 38
2.4 Bayesian Probability 39
2.5 Odds and Probability 40
2.6 Likelihood Ratio 41
2.7 Summary of Probability 42
MOSTLY ABOUT STATISTICS 44
3.1 Chi-Square for 2x2 Tables 44
3.2 McNemar Test 49
3.3 Kappa 50
3.4 Description of a Population: Use of the Standard Deviation 51
3.5 Meaning of the Standard Deviation: The Normal Distribution 55
3.6 The Difference Between Standard Deviation and Standard Error 57
3.7 Standard Error of the Difference Between Two Means 61
3.8 Z Scores and the Standardized Normal Distribution 62
3.9 The t Statistic 66
3.10 Sample Values and Population Values Revisited 67
3.11 A Question of Confidence 68
3.12 Confidence Limits and Confidence Intervals 70
3.13 Degrees of Freedom 71
3.14 Confidence Intervals for Proportions 72
3.15 Confidence Intervals Around the Difference Between Two Means 73
3.16 Comparisons Between Two Groups 75
3.17 Z-Test for Comparing Two Proportions 75
3.18 t-Test for the Difference Between Means of Two Independent Groups: Principles 77
3.19 How to Do a t-Test: An Example 79
3.20 Matched Pair t-Test 81
3.21 When Not to Do a Lot of t-Tests: The Problem of Multiple Tests of Significance 82
3.22 Analysis of Variance: Comparison Among Several Groups 84
3.23 Principles 84
3.24 Bonferroni Procedure: An Approach to Making Multiple Comparisons 87
3.25 Analysis of Variance When There Are Two Independent Variables: The Two- Factor ANOVA 89
3.26 Interaction Between Two Independent Variables 90
3.27 Example of a Two-Way ANOVA 91
3.28 Kruskal–Wallis Test to Compare Several Groups 92
3.29 Association and Causation: The Correlation Coefficient 93
3.30 How High Is High? 94
3.31 Causal Pathways 94
3.32 Regression 97
3.33 The Connection Between Linear Regression and the Correlation Coefficient 99
3.34 Multiple Linear Regression 99
3.35 Summary So Far 101
MOSTLY ABOUT EPIDEMIOLOGY 102
4.1 The Uses of Epidemiology 102
4.2 Some Epidemiologic Concepts: Mortality Rates 103
4.3 Age- Adjusted Rates 105
4.4 Incidence and Prevalence Rates 107
4.5 Standardized Mortality Ratio 109
4.6 Person- Years of Observation 109
4.7 Dependent and Independent Variables 111
4.8 Types of Studies 111
4.9 Cross-Sectional Versus Longitudinal Looks at Data 112
4.10 Measures of Relative Risk: Inferences From Prospective Studies: the Framingham Study 116
4.11 Calculation of Relative Risk from Prospective Studies 118
4.12 Odds Ratio: Estimate of Relative Risk from Case- Control Studies 119
4.13 Attributable Risk 122
4.14 Response Bias 124
4.15 Confounding Variables 126
4.16 Matching 127
4.17 Multiple Logistic Regression 128
4.18 Confounding By Indication 131
4.19 Survival Analysis: Life Table Methods 132
4.20 Cox Proportional Hazards Model 135
4.21 Selecting Variables For Multivariate Models 137
4.22 Interactions: Additive and Multiplicative Models 139
MOSTLY ABOUT SCREENING 144
5.1 Sensitivity, Specificity, and Related Concepts 144
5.2 Cutoff Point and Its Effects on Sensitivity and Specificity 151
MOSTLY ABOUT CLINICAL TRIALS 155
6.1 Features of Randomized Clinical Trials 155
6.2 Purposes of Randomization 157
6.3 How to Perform Randomized Assignment 158
6.4 Two-Tailed Tests Versus One-Tailed Test 159
6.5 Clinical Trial as “Gold Standard” 160
6.6 Regression Toward the Mean 161
6.7 Intention-to-Treat Analysis 164
6.8 How Large Should the Clinical Trial Be? 165
6.9 What Is Involved in Sample Size Calculation? 167
6.10 How to Calculate Sample Size for the Difference Between Two Proportions 171
6.11 How to Calculate Sample Size for Testing the Difference Between Two Means 172
MOSTLY ABOUT QUALITY OF LIFE 174
7.1 Scale Construction 175
7.2 Reliability 175
7.3 Validity 177
7.4 Responsiveness 178
7.5 Some Potential Pitfalls 180
MOSTLY ABOUT GENETIC EPIDEMIOLOGY 183
8.1 A New Scientific Era 183
8.2 Overview of Genetic Epidemiology 184
8.3 Twin Studies 185
8.4 Linkage and Association Studies 187
8.5 LOD Score: Linkage Statistic 190
8.6 Association Studies 191
8.7 Transmission Disequilibrium Tests (TDT) 193
8.8 Some Additional Concepts and Complexities of Genetic Studies 197
RESEARCH ETHICS AND STATISTICS 200
9.1 What does statistics have to do with it? 200
9.2 Protection of Human Research Subjects 201
9.3 Informed Consent 203
9.4 Equipoise 205
9.5 Research Integrity 205
9.6 Authorship policies 206
9.7 Data and Safety Monitoring Boards 207
9.8 Summary 207
Postscript: A FEW PARTING COMMENTS ON THE IMPACT OF EPIDEMIOLOGY ON HUMAN LIVES 208
Appendix A CRITICAL VALUES OF CHI-SQUARE, Z, AND t 210
Appendix B: FISHER’S EXACT TEST 211
Appendix C: KRUSKAL – WALLIS NONPARAMETRIC TEST TO COMPARE SEVERAL GROUPS 213
Appendix D: HOW TO CALCULATE A CORRELATION COEFFICIENT 215
Appendix E: AGE- ADJUSTMENT 217
Appendix F: CONFIDENCE LIMITS ON ODDS RATIOS 220
Appendix G: “J” OR “U” SHAPED RELATIONSHIP BETWEEN TWO VARIABLES 221
Appendix H: DETERMINING APPROPRIATENESS OF CHANGE SCORES 224
Appendix I: GENETIC PRINCIPLES 228
REFERENCES 234
SUGGESTED READINGS 239
INDEX 243

Postscript
A FEW PARTING COMMENTS ON THE IMPACT OF EPIDEMIOLOGY ON HUMAN LIVES (p. 197-198)

Ten years ago a woman with breast cancer would be likely to have a radical mastectomy, which in addition to removal of the breast and the resulting disfigurement, would also include removal of much of the muscle wall in her chest and leave her incapacitated in many ways. Today, hardly anyone gets a radical mastectomy and many don't even get a modified mastectomy, but, depending on the cancer, may get a lumpectomy which just removes the lump, leaving the breast intact. Years ago, no one paid much attention to radon, an inert gas released from the soil and dissipated through foundation cracks into homes. Now it is recognized as a leading cause of lung cancer. The role of nutrition in prevention of disease was not recognized by the scientific community. In fact, people who believed in the importance of nutrients in the cause and cure of disease were thought to be faddists, just a bit nutty. Now it is frequently the subject of articles, books, and news items, and substantial sums of research monies are invested in nutritional studies. Such studies influence legislation, as for example the regulations that processed foods must have standard labeling, easily understood by the public at large, of the fat content of the food as well as of sodium, vitamins, and other nutrients. All this has an impact on the changing eating habits of the population, as well as on the economics of the food industry.

In the health field changes in treatment, prevention, and prevailing knowledge come about when there is a confluence of circumstances: new information is acquired to supplant existing theories, there is dissemination of this information to the scientific community and to the public at large, and there is the appropriate psychological, economic, and political climate that would welcome the adoption of the new approaches. Epidemiology plays a major role by providing the methods by which new scientific knowledge is acquired. Often, the first clues to causality come long before a biological mechanism is known. Around 1850 in London, Dr. John Snow, dismayed at the suffering and deaths caused by epidemics of cholera, carefully studied reports of such epidemics and noted that cholera was much more likely to occur in certain parts of London than in other parts. He mapped the places where cholera was rampant and where it was less so, and he noted that houses supplied with water by one company, the Southwark and Vauxhall Company, had many more cases of cholera than those supplied by another company. He also knew that the Vauxhall Company used as its source an area heavily contaminated by sewage. Snow insisted that the city close the pump supplying the contaminated water, known as the Broad Street Pump. They did so and cholera abated. All this was 25 years before anyone isolated the cholera bacillus and long before people accepted the notion that disease could be spread by water. In modern times, the AIDS epidemic is one where the method of spread was identified before the infectious agent, the HIV virus, was known.

Epidemiologic techniques have been increasingly applied to chronic diseases, which differ from infectious diseases in that they may persist for a long time (whereas infections usually either kill quickly or are cured quickly) and also usually have multiple causes, many of which are difficult to identify. Here, also, epidemiology plays a central role in identifying risk factors, such as smoking for lung cancer. Such knowledge is translated into public action before the full biological pathways are elucidated. The action takes the form of educational campaigns, anti-smoking laws, restrictions on advertisement, and other mechanisms to limit smoking. The risk factors for heart disease have been identified through classic epidemiologic studies resulting in lifestyle changes for individuals as well as public policy consequences. Chronic diseases present different and challenging problems in analysis, and new statistical techniques continue to be developed to accommodate such problems. New statistical techniques are also being developed for the special problems encountered in genetics research. Thus the field of statistic is not static and the field of epidemiology is not fixed. Both adapt and expand to deal with the changing health problems of our society.