Concise Epidemiologic Principles and Concepts: Guidelines for Clinicians and Biomedical Researchers

2014 Laurens Holmes, Jr. and Franklin Opara All rights reserved.

No part of this book may be reproduced, stored in a retrieval system, or transmitted by any means without the written permission of the author.

Published by AuthorHouse 1/21/2014

ISBN: 978-1-4918-1003-3 (sc)

ISBN: 978-1-4918-1002-6 (e)

ISBN: 978-1-4918-5798-4 (hc)

Library of Congress Control Number: 2013914852

Any people depicted in stock imagery provided by Thinkstock are models,

and such images are being used for illustrative purposes only.

Certain stock imagery © Thinkstock.

Because of the dynamic nature of the Internet, any web addresses or links contained in this book may have changed since publication and may no longer be valid. The views expressed in this work are solely those of the author and do not necessarily reflect the views of the publisher, and the publisher hereby disclaims any responsibility for them.

CONTENTS

Foreword

Kirk Dabney and Richard Bowen

Preface

Laurens Holmes, Jr., and Franklin Opara

Acknowledgements

Section 1 — Research Methods

Chapter I

Research Process: Overview

Laurens Holmes, Jr., and Franklin Opara

Introduction

Structure and Function of Research

Objective of Study/Research Purpose

Research Questions and Study Hypotheses

Primary Versus Secondary Outcomes

Summary

Questions for Discussion

Chapter II

Basic Principles of Biomedical and Clinical Research

Laurens Holmes, Jr., and Franklin Opara

Introduction

Study Subjects

Sampling

Generalization

Sample Size and Power Estimations

Screening (Detection) and Diagnostic (Confirmation) Tests

Disease Screening: Principles, Advantages, and Limitations

Disease Screening: Diagnostic Tests and Clinical Reasoning

Balancing Benefits and Harmful Effects in Medicine

Summary

Questions for Discussion

Chapter III

Designing and Conducting Clinical & Biomedical Research

Laurens Holmes, Jr., and Franklin Opara

Introduction

Study Conceptualization

Research Question

Literature Review (Systematic Review): Background or Introduction

Conducting the Study and Protocol Implementation

Data Collection, Management, and Analysis

Summary

Questions for Discussion

Section 2 — Epidemiologic Principles and Concepts

Chapter IV

Epidemiologic Principles and Measure of Disease Occurrence and Association

Laurens Holmes, Jr.

Introduction

Role of Epidemiology in Clinical and Public Health Research

The History and Modern Concept of Epidemiology

Models of Disease Causation

Measures of Disease Frequency, Occurrence, and Association

Measures of Disease Association

Types of Rates (Crude, Specific, Adjusted)

Measures of Disease Comparison/Effect

Sources of Epidemiologic Data

Summary

Questions for Discussion

Chapter V

Epidemiologic Study Designs and Ecologic Studies

Laurens Holmes, Jr.

Introduction

Nonexperimental (Observational) Versus Experimental Design

Nonexperimental Ecologic Study Design and Conduct

Statistical Analysis in Ecologic Design

Summary

Questions for Discussion

Chapter VI

Cross-Sectional and Case-Control Designs

Laurens Holmes, Jr

Introduction

Cross-Sectional Design

Summary

Questions for Discussion

Chapter VII

Cohort Studies

Laurens Holmes, Jr.

Introduction

Cohort Designs

Rate Ratio Estimation in Cohort Study

Summary

Questions for Discussion

Chapter VIII

Clinical Trials (Experimental Designs)

Laurens Holmes, Jr.

Introduction

Phases of Clinical Trials

Types of Clinical Trial Designs and Statistical Inference

Elements of a Clinical Trial

The Conceptualization and Conduct of a Clinical Trial

Example of a Clinical Trial

Summary

Questions for Discussion

Chapter IX

Causal Inference in Clinical Research and Meta-Analysis

Laurens Holmes, Jr.

Introduction

Critique of Randomized Clinical Trials

Special Consideration: Critical Appraisal of Public Health/Epidemiologic Research

Quantitative Evidence Synthesis (Applied Meta-Analysis)

Statistical/Analytic Methods

Fixed Effects Model: Mantel-Haenszel and Peto

Random Effects Models: DerSimonian-Laird

Random Error and Precision

Rothman’s Component Cause Model (Causal Pies)

Summary

Questions for Discussion

Chapter X

Perspectives in Epidemiology: Advances and the Future

Laurens Holmes, Jr., and Franklin Opara

Introduction

Clinical Epidemiology

Infectious Disease Epidemiology

Molecular and Genetic Epidemiology

Cancer Epidemiology

Chronic Disease and Cardiovascular Epidemiology

Epidemiology and Health Policy Formulation

Health Disparities Epidemiology

Epidemiology: Future, Challenges and Opportunity

Summary

Questions for Discussion

About the Authors

Laurens Holmes, Jr

Franklin Opara

Endnotes

Epidemiologic design of cohort studies. Prospective cohort studies remain the hallmark of incidence data generation from non-experimental design.

We dedicate this book to our families (Holmes & Opara),

Palmer Beasly, MD, MPH (Dean Emeritus, UTSPH), in memoriam, and

James Steele, DVM, MPH (Retired Assistant US Surgeon General & Professor Emeritus, UTSPH)

FOREWORD

Clinical medicine and surgery continue to make advances by means of evidence judged to be objectively drawn from the care of individual patients. The natural observation of individuals remains the basis for our researchable questions’ formulation and the subsequent hypothesis testing. The effectiveness of evidence-based medicine or surgery is dependent on how critical we are in evaluating evidence in order to inform our practice. However, these evaluations, no matter how objective, are never absolute; rather they are probabilistic, as we will never know with absolute certainty how to treat a future patient who was not a part of our study. Despite the obstacles facing us today in attempting to provide an objective evaluation of our patients, since all of our decisions are based on judgment of some evidence, we have progressed from relying on expert opinion to the body of evidence accumulated from randomized, controlled clinical trials, as well as cohort investigations, prospective and retrospective.

Conducting a clinical trial yields more reliable and valid evidence from the data relative to nonexperimental or observational designs; however, though termed the gold standard, its validity depends on how well it is designed and conducted prior to outcome data collection, analysis, results, interpretation, and dissemination. The designs and techniques used to draw statistical inferences are often beyond the average clinician’s understanding. A text that brings study conceptualization, hypothesis formulation, design, conduct, and analysis and interpretation of the results is long overdue and highly anticipated. Epidemiology is involved with design process, which is very essential, since no amount of statistical modeling, no matter how sophisticated, can remove the error of design.

The text, Concise Epidemiologic Principles and Concepts for Clinicians and Biomedical Researchers has filled this gap, not only in the way complex designs are explained but in the simplification of statistical concepts that had never been explained in such a way before. This text has been prepared intentionally to be at the rudimentary level, so as to benefit clinicians who lack a sophisticated mathematical background or previous advanced knowledge of epidemiology, as well as biomedical researchers who may want to conduct clinical research and consumers of research products, who may benefit from the design process explained in this book. It is with this expectation and enthusiasm that we recommend this text to clinicians in all fields of clinical and biomedical research. The examples provided by the authors to simplify designs and research methods are familiar to orthopedic surgeons, as well as clinicians in other specialties of medicine and surgery.

Though statistical inference is essential in our application of the research findings to clinical decision- making regarding the care of our patients, it alone, without clinical relevance or importance, can be very misleading or even meaningless. The authors have attempted to deemphasize p value in the interpretation of epidemiologic or research findings by stressing the importance of confidence intervals, which allow for the quantification of evidence. For example, a large study, due to a large sample size that minimizes variability, may show a statistically significant difference which in reality is too insignificant to warrant any clinical importance. In contrast, the results of a small study, such as is frequently seen in clinical trials or surgical research, may have a large effect on clinical relevance but not be statistically significant at (p > 0.05). Thus, without considering the magnitude of the effect size with the confidence interval, we tend to regard these studies as negative findings, which is erroneous, since absence of evidence, based simply on an arbitrary significance level of 5 percent, does not necessarily mean evidence of absence.¹ In effect, clinical research results cannot be adequately interpreted without considering the biologic and clinical significance of the data before the statistical stability of the findings (p value and 95 percent confidence interval), since p value, as observed by the authors, merely reflects the size of the study and not the measure of evidence.

In recommending this text, it is our hope that this book will benefit clinicians, research fellows, clinical fellows, postdoctoral students in biomedical and clinical settings, nurses, clinical research coordinators, physical therapists, and all those involved in designing and conducting clinical research and analyzing research data for statistical and clinical relevance. Convincingly, knowledge gained from this text will lead to improvement of patient care through well-conceptualized research. Therefore, with the knowledge that no book is complete, no matter its content or volume, especially a book of this nature, which is prepared to guide clinicians on design, conduct, analysis, and interpretation of findings, we contend this book will benefit clinicians who are interested in applying appropriate design to research conduct.

Finally, we are optimistic that this book will bridge the gap between knowledge and practice of clinical research, especially for clinicians in a busy practice who are passionate about making a difference in their patients’ care through research and education.

Kirk Dabney, MD, MHCDS

Associate Director, Cerebral Palsy Program and Clinical Director, Health Equity & Inclusion Office, Nemours/ A.I. duPont Hospital for Children, Wilmington, Delaware

Richard Bowen, MD

Former Chairman, Orthopedic Department, A.I. duPont Hospital for Children, Wilmington, Delaware

PREFACE

Epidemiology as basic science of public health is to the public or society what clinical medicine is to the patient.

—L. Holmes, Jr. (Reality in Design Process & Statistical Modeling in Clinical & Biomedical Research, CCHS, 2012)

Appropriate knowledge of research conceptualization, design, and statistical inference is essential for conducting clinical and biomedical research. This knowledge is acquired through the understanding of epidemiologic/observational (nonexperimental) and experimental designs and the choice of the appropriate test statistic for statistical inference. However, regardless of how sophisticated the statistical technique employed for statistical inference is, study conceptualization and design are the building blocks of valid scientific evidence. Since clinical research is performed to improve patients’ care, it remains relevant to assess not only the statistical significance but the clinical and biologic importance of the findings, for clinical decision-making in the care of an individual patient. Therefore, the aim of this book is to provide clinicians, biomedical researchers, graduate students in research methodology, students of public health, and all those involved in clinical/biomedical research with a simplified but concise overview of the principles and practice of epidemiology. In addition, the authors stress common flaws in the conduct, analysis, and interpretation of epidemiologic study.

Figure i.1. Evidence discovery in biomedical and clinical research

Valid and reliable scientific research is that which considers the following elements in arriving at the truth from the data:

a)   biological relevance,

b)   clinical importance, and

c)   statistical stability (statistical inference based on the p value and the 90, 95, 99 percent confidence interval).

The interpretation of results of new research must rely on factual association or effect and the alternative explanation, namely the following:

a)   systematic error,

b)   confounding, and

c)   random error.

Therefore, unless these perspectives are disentangled, the results from any given research cannot be considered reliable. However, even with this disentanglement, all study findings remain inconclusive with some degree of uncertainty.

This book presents a concise guide on how to conduct clinical research—mainly, research question formulation, study implementation, hypothesis testing using appropriate test statistics to analyze the data, and results interpretation. In so doing, it attempts to illustrate the basic concepts used in study conceptualization, epidemiologic design, and appropriate test statistics for statistical inference from the data. Therefore, though statistical inference is emphasized throughout the presentation in this text, equal emphasis is placed on clinical relevance or importance and biological relevance in the interpretation of the study results. A study could be statistically significant but biologically and clinically irrelevant, since the statistical stability of a study does not rule out bias and confounding. The p value is deemphasized, while the use of effect size or magnitude and confidence intervals in the interpretation of results for application in clinical decision-making is recommended. The use of p value could lead to an erroneous interpretation of the effectiveness of a treatment. For example, studies with large sample sizes and very little or insignificant effects of no clinical importance may be statistically significant, while studies with small samples though a large magnitude of effects are labeled negative result.² Such results are due to low statistical power and increasing variability, hence the inability to pass the arbitrary litmus test of the 5 percent significance level.

What is epidemiology?

Epidemiologic investigation and practice is as old as the history of modern medicine. It dates back to Hippocrates (circa 2,400 years ago). In recommending the appropriate practice of medicine, Hippocrates appealed to the physicians’ ability to understand the role of environmental factors in predisposition to disease and health in the community. During the Middle Ages and the Renaissance, epidemiologic principles continued to influence the practice of medicine, as demonstrated in De Morbis Artificum (1713) by Ramazinni and the works on scrotal cancer in relation to chimney sweeps by Percival Pott in 1775.

With the works of John Snow, a British physician (1854), on cholera mortality in London, the era of scientific epidemiology began. By examining the distribution/pattern of mortality and cholera in London, Snow postulated that cholera was caused by contaminated water.

What is epidemiology today?

There are several definitions of epidemiology, but a practical definition is necessary for the understanding of this human science. Epidemiology is the basic science of public health. The objective of this discipline is to assess the distribution and determinants of disease, disabilities, injuries, natural disasters (tsunamis, hurricanes, tornados, and earthquakes) and health-related events at the population level. Epidemiologic investigation or research focuses on a specific population. The basic issue is to assess the groups of people at higher risk: women, children, men, pregnant women, teenagers, whites, African Americans, Hispanics, Asians, poor, affluent, gay, lesbians, married, single, older individuals, etc. Epidemiology also examines how the frequency of the disease or the event of interest changes over time. In addition, epidemiology examines the variation of the disease of interest from place to place. Simply, descriptive epidemiology attempts to address the distribution of disease with respect to who, when, and where. For example, cancer epidemiologists attempt to describe the occurrence of prostate cancer by observing the differences in populations by age, socioeconomic status, occupation, geographic locale, race/ethnicity, etc. Epidemiology also attempts to address the association between the disease and exposure. For example, why are some men at high risk for prostate cancer? Does race/ethnicity increase the risk for prostate cancer? Simply, is the association causal or spurious? This process involves the effort to determine whether a factor (exposure) is associated with the disease (outcome). In the example with prostate cancer, such exposure includes a high-fat diet, race/ethnicity, advancing age, pesticides, family history of prostate cancer, and so on. Whether or not the association is factual or a result of chance remains the focus of epidemiologic research. The questions to be raised are as follows: Is prostate cancer associated with pesticides? Does pesticide cause prostate cancer?

Epidemiology often goes beyond disease-exposure association or relationship to establish causal association. In this process of causal inference, it depends on certain criteria, one of which is the strength or magnitude of association, leading to the recommendation of preventive measures. However, complete knowledge of the causal mechanism is not necessary prior to preventive measures for disease control. Further, findings from epidemiologic research facilitate the prioritization of health issues and the development and implementation of intervention programs for disease control and health promotion.

This book is conceptually organized in two sections. Section 1 deals with research methods and section 2 epidemiologic designs, as well as causal inference and perspectives in epidemiology. Throughout this book, attempts are made to describe the research methods and nonexperimental as well as experimental designs. Section 1 comprises research methods with an attempt to describe the following:

260339.png    Research objectives and purposes

260341.png    Research questions

260343.png    Hypothesis statements: null and alternative

260345.png    Rationales for research, clinical reasoning, and diagnostic tests

260347.png    Study conceptualization and conduct—research question, data collection, data management, hypothesis testing, data analysis

Section 2 comprises the epidemiologic study designs with an attempt to describe the basic notion of epidemiology and the designs used in clinical research:

260359.png    The notion of epidemiology and the measures of disease occurrence and frequency and the measure of disease association

260362.png    Ecologic and cross-sectional designs

260364.png    Case-control studies

260366.png    Cohort studies: prospective, retrospective, and ambidirectional

260368.png    Clinical trials or experimental designs

260370.png    Quantitative evidence synthesis (QES), meta-analysis, scientific study appraisal, and causal inference

260372.png    Epidemiologic perspectives: advances and the future

Section 1 has three chapters. The first two chapters deal with the basic descriptions of scientific research at the clinical and population levels and how the knowledge gained from the population could be applied to the understanding of individual patients in the future. The attempt is made in these two chapters to discuss clinical reasoning and the use of diagnostic tests (sensitivity and specificity) in clinical decision-making. The notions, numbers needed to treat (NNT), and numbers needed to harm (NNH) are discussed later in the chapter on causal inference. The last chapter in this section, chapter 3, delves into clinical research conceptualization, design involving subject recruitment, variable ascertainment, data collection, data management, data analysis, and the outline of the research proposal.

In section 2, epidemiologic principles and methods are presented with the intent to stress the importance of a careful design in conducting clinical and biomedical research. Epidemiology remains the basic science of clinical medicine and public health that deals with disease and injury distributions and determinants and the application of this knowledge to the control and prevention of disease, disabilities, injuries, and related health events at the population level. Depending on the research question and whether or not the outcome (disease or event of interest) has occurred prior to the commencement of the study or the investigator assigns subjects to treatment or control, an appropriate design is selected for the clinical research. The measures of effects or point estimates are discussed with concrete examples to illustrate the application of epidemiologic principles in arriving at a standard valid result. Designs are illustrated with flow charts, figures, and boxes for distinctions and similarities. The hierarchy of study design is demonstrated with randomized clinical trial (RCT) and the associated meta-analysis as the design that yields the most reliable and valid evidence from data. Though RCTs are considered the gold standard of clinical research, it is sometimes not feasible to use this design because of ethical considerations, hence the alternative need for prospective cohort design.

Interpreting research findings is equally as essential as conducting the study itself. Interpretation of research findings must be informative and constructive in order to identify future research needs. A research result cannot be considered valid unless we disentangle the role of bias and confounding from a statistically significant finding, as a study can be statistically significant and yet driven by measurement, selection, and information bias as well as confounding. While my background (LH) in basic medical sciences and clinical medicine (internal medicine) allows me to appreciate the importance of biologic and clinical relevance in the interpretation of research findings, biostatisticians without similar training must look beyond random variation (p value and confidence interval) in the interpretation and utilization of clinical and biomedical research findings. Therefore, quantifying the random error with p value (a meaningful null hypothesis with a strong case against the null hypothesis requires the use of significance level) without a confidence interval deprives the reader of the ability to assess the clinical importance of the range of values in the interval. Using Fisher’s arbitrary p value cutoff point for type I error (alpha level) tolerance, a p value of 0.05 need not provide strong evidence against the null hypothesis, but p less than 0.0001 does.³ The precise p value should be presented, without reference to arbitrary thresholds. Therefore, results of clinical and biomedical research should not be presented as significant or nonsignificant but should be interpreted in the context of the type of study and other available evidence. Secondly, systematic error and confounding should always be considered for findings with low p values, as well as the potentials for effect measure modifier (if any) in the explanation of the results. Neyman and Pearson describe their accurate observation:

No test based upon a theory of probability can by itself provide any valuable evidence of the truth or falsehood of a hypothesis. But we may look at the purpose of tests from another viewpoint. Without hoping to know whether each separate hypothesis is true or false, we may search for rules to govern our behavior with regard to them, in following which we insure that, in the long run of experience, we shall not often be wrong.⁴

Epidemiology is an ever-changing discipline. Authors have consulted with data judged to be accurate at the moment of the presentation of these materials for publication. However, due to rapid changes in risk factors identification and biomarkers of disease, the limitations of human knowledge and the possibity of errors, the authors wish to be insulated from any responsibility due to error arising from the use of this text. Therefore readers are advised to consult with other sources of similar data for the confirmation of the information therein.

ACKNOWLEDGEMENTS

In preparing this book, so many people contributed directly or indirectly to the materials provided here. In order to make this book practical, we used data collected from different studies. We wish to express sincere gratitude to those who permitted the use of their data to illustrate the design techniques used in this book.

The attempt to create a simplified book in epidemiology remains challenging because of variability in the epidemiologic reasoning of those who require such materials. The simplification, so to speak, of the design process in evidence in clinical research came from my interaction (LH) with research fellows at the Nemours Orthopedic Department, who worked with us and gave us the reason to write this book. They are Drs. Durga K.; Ali, M.; Joo, S.; Palocaren, T; Haumont, T.; Cornes, M.J.; Tahbet, A.; Atanda, A; Connor, J; Oto, M.; Kadhim, M.; and Karatas, A. The clinical research fellows and surgical residents in the orthopedic department also motivated the preparation of this text. Thank you for your interest in the evidence-based journal club.

My colleagues at the Nemours Orthopedic Department, Nemours Center for Childhood Cancer