Handbook for Clinical Research Design

Statistics and Implementation 1st

Edition Flora Hammond
Visit to download the full and correct content document:
https://ebookmeta.com/product/handbook-for-clinical-research-design-statistics-and-i
mplementation-1st-edition-flora-hammond/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Handbook of Hormones: Comparative Endocrinology for

Basic and Clinical Research, 2nd Edition Hironori Ando

https://ebookmeta.com/product/handbook-of-hormones-comparative-
endocrinology-for-basic-and-clinical-research-2nd-edition-
hironori-ando/

Research Methods The Key Concepts Michael Hammond

https://ebookmeta.com/product/research-methods-the-key-concepts-
michael-hammond/

Research Handbook on Implementation of Human Rights in

Practice 1st Edition Rachel Murray

https://ebookmeta.com/product/research-handbook-on-
implementation-of-human-rights-in-practice-1st-edition-rachel-
murray/

Virtual and Remote Control Tower Research Design

Development Validation and Implementation 2nd Edition
Norbert Fürstenau

https://ebookmeta.com/product/virtual-and-remote-control-tower-
research-design-development-validation-and-implementation-2nd-
edition-norbert-furstenau/
Activities for Teaching Statistics and Research Methods
1st Edition Jeffrey R. Stowell

https://ebookmeta.com/product/activities-for-teaching-statistics-
and-research-methods-1st-edition-jeffrey-r-stowell/

Using IBM SPSS Statistics for Research Methods and

Social Science Statistics 7th Edition William E Wagner

https://ebookmeta.com/product/using-ibm-spss-statistics-for-
research-methods-and-social-science-statistics-7th-edition-
william-e-wagner/

Bathrooms and Sanitation Principles Design

Implementation 1st Edition Sibylle Kramer

https://ebookmeta.com/product/bathrooms-and-sanitation-
principles-design-implementation-1st-edition-sibylle-kramer/

Research Methods The Key Concepts 2nd Edition Michael

Hammond Jerry Wellington

https://ebookmeta.com/product/research-methods-the-key-
concepts-2nd-edition-michael-hammond-jerry-wellington/

Fundamentals of Statistics for Aviation Research 1st

Edition Michael A. Gallo

https://ebookmeta.com/product/fundamentals-of-statistics-for-
aviation-research-1st-edition-michael-a-gallo/
Handbook for

Handbook for Clinical Research

Clinical Research
Design, Statistics, and Implementation
Flora M. Hammond, MD • James F. Malec, PhD
Todd G. Nick, PhD • Ralph M. Buschbacher, MD
Handbook for
W ith over 80 information-packed chapters, Handbook for Clinical Research delivers
the practical insights and expert tips necessary for successful research design, analysis,
and implementation. Using clear language and an accessible bullet point format, the
Clinical Research
authors present the knowledge and expertise developed over time and traditionally shared from
mentor to mentee and colleague to colleague. Organized for quick access to key topics and replete
Design, Statistics, and Implementation
with practical examples, the book describes a variety of research designs and statistical methods and
explains how to choose the best design for a particular project. Research implementation, includ-
ing regulatory issues and grant writing, is also covered.
The book opens with a section on the basics of research design, discussing the many ways in which
Flora M. Hammond

Design, Statistics, and Implementation

studies can be organized, executed, and evaluated. The second section is devoted to statistics; it
explains how to choose the correct statistical approach and reviews the varieties of data types,
James F. Malec
Todd G. Nick
descriptive and inferential statistics, methods for demonstrating associations, hypothesis testing and
prediction, specialized methods, and considerations in epidemiological studies and measure
construction. The third section covers implementation, including how to develop a grant
application step by step, the project budget, and the nuts and bolts of the timely and successful
completion of a research project and documentation of findings: procedural manuals and case
Ralph M. Buschbacher
report forms; collecting, managing and securing data; operational structure and ongoing monitor-
ing and evaluation; and ethical and regulatory concerns in research with human subjects.
With a concise presentation of the essentials for successful research, the Handbook for Clinical
Research is a valuable addition to the library of any student, researcher, professional, or clinician
interested in expanding the knowledge base of his or her field.

Hammond • Malec • Nick • Buschbacher

Key Features:
■■ Delivers the essential elements, practical insights, and trade secrets for ensuring
successful research design, analysis, and implementation
■■ Presents the nuts and bolts of statistical analysis
■■ Organized for quick access to a wealth of information
■■ Replete with practical examples of successful research designs—from single case
designs to meta-analysis—and how to achieve them
■■ Addresses research implementation including regulatory issues and grant writing

Recommended
Shelving Category:
Medical

11 West 42nd Street

New York, NY 10036
www.demosmedical.com
9 781936 287543
 Handbook for
Clinical Research

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd i 8/11/2014 10:59:15 AM

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd ii 8/11/2014 10:59:15 AM
 Handbook for
Clinical Research
Design, Statistics, and Implementation

Editors
Flora M. Hammond, MD
Chair, Department of Physical Medicine and Rehabilitation
Covalt Professor of Physical Medicine and Rehabilitation
Indiana University School of Medicine
Chief of Medical Affairs
Rehabilitation Hospital of Indiana
Indianapolis, Indiana

James F. Malec, PhD

Professor and Research Director, Department of Physical Medicine and Rehabilitation
Indiana University School of Medicine and
Rehabilitation Hospital of Indiana
Indianapolis, Indiana
Professor Emeritus
Mayo Clinic

Todd G. Nick, PhD

Director, Biostatistics Program
Professor, Department of Pediatrics
University of Arkansas for Medical Sciences
Little Rock, Arkansas

Ralph M. Buschbacher, MD
Clinical Professor, Department of Physical Medicine and Rehabilitation
Indiana University School of Medicine
Indianapolis, Indiana

New York

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd iii 8/11/2014 10:59:15 AM

 Visit our website at www.demosmedical.com

ISBN: 978-1-9362-8754-3
e-book ISBN: 978-1-6170-5099-2

Acquisitions Editor: Beth Barry

Compositor: Newgen Knowledge Works

© 2015 Demos Medical Publishing, LLC. All rights reserved. This book is protected by copyright. No part of it may be
reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the publisher.

Medicine is an ever-changing science. Research and clinical experience are continually expanding our knowledge, in par-
ticular our understanding of proper treatment and drug therapy. The authors, editors, and publisher have made every effort
to ensure that all information in this book is in accordance with the state of knowledge at the time of production of the book.
Nevertheless, the authors, editors, and publisher are not responsible for errors or omissions or for any consequences from
application of the information in this book and make no warranty, expressed or implied, with respect to the contents of the
publication. Every reader should examine carefully the package inserts accompanying each drug and should carefully check
whether the dosage schedules mentioned therein or the contraindications stated by the manufacturer differ from the state-
ments made in this book. Such examination is particularly important with drugs that are either rarely used or have been newly
released on the market.

Library of Congress Cataloging-in-Publication Data

Handbook for clinical research : design, statistics, and implementation / editors, Flora M. Hammond, James F. Malec, Todd
G. Nick, Ralph M. Buschbacher.
p. ; cm.
Includes bibliographical references and index.
ISBN 978-1-936287-54-3—ISBN (invalid) 978-1-61705-099-2 (e-book)
I. Hammond, Flora, editor. II. Malec, James F., editor. III. Nick, Todd, editor. IV. Buschbacher, Ralph M., editor.
[DNLM: 1. Biomedical Research. 2. Clinical Trials as Topic. 3. Epidemiologic Research Design. 4. Statistics as Topic.
W 20.5]
R853.C55
610.72’4—dc23
2014015700

Special discounts on bulk quantities of Demos Medical Publishing books are available to corporations, professional asso-
ciations, pharmaceutical companies, health care organizations, and other qualifying groups. For details, please contact:
Special Sales Department
Demos Medical Publishing, LLC
11 West 42nd Street, 15th Floor
New York, NY 10036
Phone: 800-532-8663 or 212-683-0072
Fax: 212-941-7842
E-mail: specialsales@demosmedical.com

Printed in the United States of America by Gasch Printing.

14 15 16 17 / 5 4 3 2 1

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd iv 8/11/2014 10:59:16 AM

 Contents

Contributors ix
Preface xiii
Share Handbook for Clinical Research: Design, 10. Choice of Control Groups in Treatment
Statistics, and Implementation Studies 37
Sureyya S. Dikmen, PhD
PART I: DESIGN
11. Randomization 40
Emilia Bagiella, PhD
1. Development and Testing of Treatments 3
Eric Lenze, MD and John Whyte, MD, PhD 12. Special Issues in Randomized Controlled
Trials 43
2. Qualitative Research 9
Grace Handler and Michael L. Boninger, MD
Elena Gillespie, PhD
13. Secondary Data Analysis 46
3. Single-Case Experimental Designs 13
Kenneth J. Ottenbacher, PhD, OTR,
Robyn L. Tate, MPsychol, PhD and Michael
James E. Graham, PhD, DC, and
Perdices, MA (ClinNeuropsychology), PhD
Amol Karmarkar, PhD, MPH, OTR
4. Studies of Associations 18
14. Scoping Study 50
Michael Schönberger, PhD, Dipl.-Psych.
Xinsheng Cai, PhD
5. Observational Studies: Retrospective Versus
15. Systematic Reviews 53
Prospective 23
Naomi Lynn Gerber, MD and
Brian Keogh, Jr, MD, and
Xinsheng Cai, PhD
Katherine W. Stenson, MD
16. Meta-Analysis 57
6. Historical Controls 27
Xinsheng Cai, PhD
Whitney Pratt, MD, PhD and
Katherine W. Stenson, MD 17. Recommendations for Reporting Research
Studies 62
7. Subject as Own Control 30
Ronald T. Seel, PhD
Lisa A. Lombard, MD
18. Developing and Evaluating Systematic Reviews
8. Longitudinal Cohort Versus
and Practice Guidelines 68
Cross-Sectional Cohort Studies 32
Ronald T. Seel, PhD
Zachary Timothy Glynn Siegel, MD and
Katherine W. Stenson, MD
9. Survey Research 34
Amanda Leigh Harrington, MD

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd v 8/11/2014 10:59:16 AM

 vi CONTENTS

PART II: STATISTICS 37. Significance Tests: Categorical Data 152

Xinyu Tang, PhD
19. Introduction 77 38. Measures of Effect Sizes for Categorical
Todd G. Nick, PhD and Christopher J. Outcomes 155
Swearingen, PhD Todd G. Nick, PhD
20. Types of Data 79 39. Logistic Regression 158
Maria S. Melguizo, MS and Marcus Wellen, MD Todd G. Nick, PhD and Chunqiao Luo, MS
21. Descriptive Statistics 84 40. Kaplan–Meier Estimator 164
Todd G. Nick, PhD Xinyu Tang, PhD
22. Data Distributions 87 41. Log-Rank Test 167
Maria S. Melguizo, MS and Marcus Wellen, MD Xinyu Tang, PhD
23. Samples and Populations 91 42. Proportional Hazards Model 169
Maria S. Melguizo, MS and Marcus Wellen, MD Xinyu Tang, PhD
24. Visual Display of Data 93 43. Sources of Error: Selection Bias, Information
Maria S. Melguizo, MS and Marcus Wellen, MD Bias, and Confounding 171
Jareen Meinzen-Derr, PhD, MPH and Laura
25. Data Cleaning 101
Smith, MPH
Jingyun Li, MS
44. Mediation Analyses 174
26. Missing Data and Imputation 105
Bin Huang, PhD and Todd G. Nick, PhD
Christopher J. Swearingen, PhD
45. Epidemiology Study: Incidence and
27. Estimation 108
Prevalence 178
Mallikarjuna Rettiganti, PhD
Jareen Meinzen-Derr, PhD, MPH and Laura
28. Hypothesis Testing 112 Smith, MPH
Mallikarjuna Rettiganti, PhD
46. Validity and Performance of Screening:
29. Sample Size and Power 115 Sensitivity, Specificity, Positive Predictive Value,
Mallikarjuna Rettiganti, PhD and Negative Predictive Value 182
Jareen Meinzen-Derr, PhD, MPH and Laura
30. Comparing Matched Samples With
Smith, MPH
Continuous-Type Outcomes: Two Groups 119
Ming Li, PhD 47. Statistical Tools for Agreement and Reliability
Studies 185
31. Comparing Independent Samples With
Shasha Bai, PhD
Continuous-Type Outcomes: Two Groups 124
Ming Li, PhD 48. Classical Test Theory 192
Jacob Kean, PhD and Jamie Reilly, PhD
32. Comparing Independent Samples for
Continuous-Type Outcomes: Three Groups or 49. Item Response Theory 195
More 129 Jacob Kean, PhD and Jamie Reilly, PhD
Ming Li, PhD
33. Correlation 133
PART III: IMPLEMENTATION
Christopher J. Swearingen, PhD
34. Simple Linear Regression 137 Grant Proposals
Christopher J. Swearingen, PhD
50. Successful Grant Applications 201
35. Multiple Linear Regression 142 Flora M. Hammond, MD
Christopher J. Swearingen, PhD
51. Sources of Research Funding 204
36. Longitudinal and Clustered Data 147 James F. Malec, PhD
Mallikarjuna Rettiganti, PhD

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd vi 8/11/2014 10:59:16 AM

 CONTENTS vii

52. Planning Grants 208 69. Participant Recruitment and Enrollment 266
Eric Lenze, MD Flora M. Hammond, MD
53. Developing the Idea With Stakeholder 70. Participant Retention 271
Input 211 Flora M. Hammond, MD
John D. Corrigan, PhD
71. Data Collection 276
54. Research Questions, Hypotheses, Aims, and Elena Gillespie, PhD and Flora M. Hammond, MD
Abstract 214
72. Case Report Forms 280
Flora M. Hammond, MD
Marybeth Whitney, BSN, CRA and Flora M.
55. Reviewing the Literature 218 Hammond, MD
Dawn Neumann, PhD and
73. Database Development 283
Edward Garay, MD, PhD
Emilia Bagiella, PhD
56 Background and Significance 222
74. Data Dictionary 286
Flora M. Hammond, MD
Emilia Bagiella, PhD
57. Preliminary Studies and Experience 224
75. Data Management 288
Flora M. Hammond, MD
Emilia Bagiella, PhD
58. Methods and Design 226
76. Plan of Operation 291
Kathleen R. Bell, MD
Angelle M. Sander, PhD
59. Types of Measures 231
77. Evaluation 295
James F. Malec, PhD
Angelle M. Sander, PhD
60. Letters of Support 234
78. Regulatory Binder and Essential
James F. Malec, PhD
Documents 298
61. Budget and Budget Justification 236 Marybeth Whitney, BSN, CRA and Flora M.
LaMoria Patterson, BS and Rob Dimmitt, MBA, BS Hammond, MD
62. Preaward Management 241 79. Adverse Events 302
LaMoria Patterson, BS and Rob Dimmitt, MBA, BS Marybeth Whitney, BSN, CRA and Flora M.
Hammond, MD
Conducting the Research
80. Protocol Deviations and Violations 307
63. Post-Award Management 245 Marybeth Whitney, BSN, CRA and Flora M.
Rob Dimmitt, MBA, BS and LaMoria Patterson, BS Hammond, MD
64. Good Clinical Practices 249 81. Data and Safety Monitoring 310
Shawn Axe, CIP and John R. Baumann, PhD Emilia Bagiella, PhD
65. Research Misconduct 252 82. Multicenter Trials 315
John R. Baumann, PhD, Shelley Bizila, MS, CIP, Flora M. Hammond, MD
CCEP, and Rebecca Hopson, BA
83. Site Monitoring and Oversight 318
66. Study Protocol 256 Marybeth Whitney, BSN, CRA and Flora M.
Flora M. Hammond, MD Hammond, MD
67. Manual of Procedures 260
Index 323
Flora M. Hammond, MD
68. Treatment Manuals 263
Kathleen R. Bell, MD

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd vii 8/11/2014 10:59:16 AM

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd viii 8/11/2014 10:59:16 AM
 Contributors

Shawn Axe, CIP Shelley Bizila, MS, CIP, CCEP

Associate Director Clinical Research Compliance Officer
Indiana University Human Subjects Office Research Integrity Officer
Office of Research Administration Office of Research Compliance
Indiana University Indiana University
Indianapolis, IN Indianapolis, IN

Emilia Bagiella, PhD Michael L. Boninger, MD

Professor UPMC Endowed Professor and Chair
Center for Biostatistics Department of Physical Medicine and Rehabilitation
Department of Health Evidence and Policy University of Pittsburgh;
Icahn School of Medicine at Mount Sinai Medical Director
New York, NY VA Center of Excellence in Wheelchairs and Related
Mobility
Shasha Bai, PhD VA Pittsburgh Health Care System
Assistant Professor Department of Veterans Affairs
Biostatistics Program Pittsburgh, PA
Department of Pediatrics
University of Arkansas for Medical Sciences Xinsheng Cai, PhD
Little Rock, AR Senior Researcher
American Institutes for Research
John R. Baumann, PhD Washington, DC
Assistant Vice President for Research Compliance
Office of Research Compliance John D. Corrigan, PhD
Office of the Vice President for Research Professor
Indiana University Department of Physical Medicine and Rehabilitation
Indianapolis, IN Director
Division of Rehabilitation Psychology
Kathleen R. Bell, MD Ohio State University;
Professor Director
Department of Rehabilitation Medicine Ohio Valley Center for Brain Injury Prevention and
University of Washington Rehabilitation
Seattle, WA Columbus, OH

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd ix 8/11/2014 10:59:16 AM

 x CONTRIBUTORS

Sureyya S. Dikmen, PhD Rebecca Hopson, BA

Professor Clinical Research Compliance Project Coordinator
Department of Rehabilitation Medicine Clinical Research Compliance Office
University of Washington Indiana University
Seattle, WA Indianapolis, IN

Rob Dimmitt, MBA, BS Bin Huang, PhD

Director of Finance Associate Professor
Indiana Clinical and Translational Sciences Institute Division of Biostatistics and Epidemiology
Indiana University Department of Pediatrics
Indianapolis, IN Cincinnati Children’s Hospital Medical Center
Cincinnati, OH
Edward Garay, MD, PhD
Resident Amol Karmarkar, PhD, MPH, OTR
Department of Physical Medicine and Rehabilitation Assistant Professor
University of Pittsburgh School of Medicine Division of Rehabilitation Sciences
Pittsburgh, PA University of Texas Medical Branch at Galveston
Galveston, TX
Naomi Lynn Gerber, MD
Professor Jacob Kean, PhD
The Center for Study of Chronic Illness and Disability Research Health Scientist
Department Health Administration and Policy VA HSR&D Center for Health Information and
George Mason University Communication
Fairfax, VA Richard L. Roudebush VA Medical Center;
Assistant Research Professor
Elena Gillespie, PhD Department of Physical Medicine and Rehabilitation
Research Coordinator Indiana University School of Medicine;
Department of Research Research Scientist
Rehabilitation Hospital of Indiana Regenstrief Institute
Indianapolis, IN Indianapolis, IN

James E. Graham, PhD, DC Brian Keogh, Jr, MD

Associate Professor Fellow
Division of Rehabilitation Sciences Division of Pain Medicine
University of Texas Medical Branch at Galveston Department of Anesthesia
Galveston, TX Emory University
Atlanta, GA
Flora M. Hammond, MD
Chair, Department of Physical Medicine and Rehabilitation Eric Lenze, MD
Covalt Professor of Physical Medicine and Rehabilitation Professor
Indiana University School of Medicine; Department of Psychiatry
Chief of Medical Affairs Washington University School of Medicine
Rehabilitation Hospital of Indiana St. Louis, MO
Indianapolis, IN
Jingyun Li, MS
Grace Handler Biostatistician
Student Intern Biostatistics Program
UPMC Rehabilitation Institute Department of Pediatrics
UPMC University of Arkansas for Medical Sciences
Pittsburgh, PA Little Rock, AR

Amanda Leigh Harrington, MD Ming Li, PhD

Assistant Professor Assistant Professor
Department of Physical Medicine and Rehabilitation Biostatistics Program
University of Pittsburgh School of Medicine; Department of Pediatrics
Director of Spinal Cord Injury Services University of Arkansas for Medical Sciences
UPMC Rehabilitation Institute Little Rock, AR
UPMC Mercy
Pittsburgh, PA

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd x 8/11/2014 10:59:16 AM

 CONTRIBUTORS xi

Lisa A. Lombard, MD LaMoria Patterson, BS

Medical Director Lead Research Administrator
Rehabilitation Hospital of Indiana; Neuroscience Administration
Assistant Professor Indiana University Health Neuroscience Center
Department of Physical Medicine and Rehabilitation Indiana University School of Medicine
Indiana University School of Medicine Indianapolis, IN
Indianapolis, IN
Michael Perdices, MA (ClinNeuropsychology), PhD
Chunqiao Luo, MS Department of Neurology
Biostatistician Royal North Shore Hospital
Biostatistics Program St. Leonards, NSW, Australia;
Department of Pediatrics Senior Clinical Lecturer
University of Arkansas for Medical Sciences Division of Psychological Medicine
Little Rock, AR Sydney Medical School–Northern
University of Sydney
James F. Malec, PhD Sydney, Australia
Professor and Research Director
Department of Physical Medicine and Rehabilitation Whitney Pratt, MD, PhD
Indiana University School of Medicine; Assistant Clinical Professor
Rehabilitation Hospital of Indiana Department of Physical Medicine and Rehabilitation
Indianapolis, IN Indiana University School of Medicine
Professor Emeritus Indianapolis, IN
Mayo Clinic
Jamie Reilly, PhD
Jareen Meinzen-Derr, PhD, MPH Assistant Professor
Associate Professor Department of Communication Sciences & Disorders
Division of Biostatistics and Epidemiology Eleanor M. Saffran Center for Cognitive Neuroscience
Department of Pediatrics Temple University
Cincinnati Children’s Hospital Medical Center Philadelphia, PA
Cincinnati, OH
Mallikarjuna Rettiganti, PhD
Maria S. Melguizo, MS Assistant Professor
Research Program Manager Biostatistics Program
Biostatistics Program Department of Pediatrics
Department of Pediatrics University of Arkansas for Medical Sciences
University of Arkansas for Medical Sciences Little Rock, AR
Little Rock, AR
Angelle M. Sander, PhD
Dawn Neumann, PhD Associate Professor
Assistant Research Professor Physical Medicine and Rehabilitation
Department of Physical Medicine and Rehabilitation Baylor College of Medicine and Harris Health System;
Indiana University School of Medicine Director and Senior Scientist
Indianapolis, IN Brain Injury Research Center
TIRR Memorial Hermann
Todd G. Nick, PhD Houston, TX
Director, Biostatistics Program
Professor Michael Schönberger, PhD, Dipl.-Psych.
Department of Pediatrics Department of Rehabilitation Psychology and
University of Arkansas for Medical Sciences Psychotherapy
Little Rock, AR Institute of Psychology
University of Freiburg
Kenneth J. Ottenbacher, PhD, OTR Freiburg, Germany;
Professor and Director School of Psychological Sciences
Division of Rehabilitation Sciences Monash University
University of Texas Medical Branch at Galveston Melbourne, Australia
Galveston, TX

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd xi 8/11/2014 10:59:16 AM

 xii CONTRIBUTORS

Ronald T. Seel, PhD Robyn L. Tate, MPsychol, PhD

Director, Brain Injury Research
Crawford Research Institute
Shepherd Center
Atlanta, GA
Zachary Timothy Glynn Siegel, MD
Department of Physical Medicine and Rehabilitation
Indiana University School of Medicine
Indianapolis, IN
Professorial Research Fellow
Rehabilitation Studies Unit
Sydney Medical School–Northern
University of Sydney
Sydney, Australia;
Kolling Institute of Medical Research
Royal North Shore Hospital
St. Leonards, NSW, Australia

Laura Smith, MPH Marcus Wellen, MD

Clinical Research Coordinator Staff Psychiatrist
Department of Pediatrics Department of Mental Health and Behavioral
Cincinnati Children’s Hospital Medical Center Medicine
Cincinnati, OH Central Texas Veterans Healthcare System
Austin, TX
Katherine W. Stenson, MD
Assistant Professor
Department of Physical Medicine and Rehabilitation Marybeth Whitney, BSN, CRA
Indiana University School of Medicine; Manager, Clinical Trials
Medical Director Department of Physical Medicine and
Spinal Cord Injury Program Rehabilitation
Rehabilitation Hospital of Indiana Carolinas HealthCare System
Indianapolis, IN Charlotte, NC

Christopher J. Swearingen, PhD

Associate Professor John Whyte, MD, PhD
Pediatric Biostatistics Director
Department of Pediatrics Moss Rehabilitation Research Institute
University of Arkansas for Medical Sciences Elkins Park, PA;
Little Rock, AR Professor
Department of Rehabilitation Medicine
Xinyu Tang, PhD Jefferson Medical College
Assistant Professor Thomas Jefferson University
Associate Director Philadelphia, PA
Biostatistics Program
Department of Pediatrics
University of Arkansas for Medical Sciences
Little Rock, AR

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd xii 8/11/2014 10:59:16 AM


We started working on this text at the encouragement
of Dr. Ralph Buschbacher to develop a quick refer-
ence guide to research primarily for rehabilitation
clinicians. Initially, we imagined a relatively thin
volume that focused on the key elements of rehabili-
tation research. However, as we began to formalize
the outline, the scope of the book rapidly grew to
the 80-plus chapters in the present volume. As we
recruited authors for the various chapters and began
initial editing, we quickly realized that the content
of these chapters was applicable not just to rehabil-
itation, but to other disciplines involved in clinical
research, eg, other medical specialties, psychology,
nursing; and the scope of the book expanded further.
The potential impact of the book also became more
apparent to us. This book gave us a chance to out-
line not only the key elements of clinical research
but also to put down in writing the “trade secrets,”
shortcuts, practical insights, and helpful hints to con-
ducting clinical research that we and the other authors
have learned over the years by word of mouth from
mentor to mentee and colleague to colleague. Our
excitement grew about developing a text that might
be a quick reference guide for clinical researchers in
any discipline—a book that could assist them in their
daily work planning and conducting research studies,
journal and grant reviews, and mentoring. We began
to imagine a book that might serve as a supplemen-
tary text to courses on research design and analysis
as well as grant writing courses and workshops, or
that might be a good primary text on research for
any clinical training program with a clinical research
component.
Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd xiii
Preface
And so, the Handbook for Clinical Research:
Design, Statistics, and Implementation took shape.
Part I covers the basics of research design: the variety
of ways in which studies can be organized to address
questions of association or causation; the appropri-
ate sequencing of studies in a particular area to move
most efficiently from demonstration of concept to a
definitive and rigorous trial; methods and appropriate
rigor of control conditions; retrospective and prospec-
tive trials; qualitative and quantitative analyses; and
methods for summarizing, evaluating, and reporting
clinical research in a particular area. Part II, statistics,
begins with a discussion of selecting the correct statis-
tical approach and when to consult a statistician. Part
II then reviews the varieties of data types; descriptive
and inferential statistics; methods for demonstrat-
ing associations, hypothesis testing, and prediction;
specialized methods, such as survival modeling; and
specialized methods for epidemiological studies and
measure construction. Part III, implementation, begins
with a number of chapters on developing successful
grant applications from planning and seeking consumer
input to developing specific sections of research grant
proposals, the project budget, and ancillary materials.
The final chapters of this section cover the nuts and
bolts of the timely and successful completion of the
research project; developing procedural manuals and
case report forms; collecting, managing, and securing
data; operational structure and ongoing monitoring
and evaluation of the project; and ethical and regula-
tory concerns in research with human subjects.
Many of the authors (including ourselves) at
first found the succinct, focused format of the book to
8/11/2014 10:59:16 AM
 xiv PREFACE

be challenging. Most of us would have found it easier
to write the traditional 30-odd-page chapter replete
with extensive referencing and interesting (at least to
us) intellectual alleyways. However, all of the authors
rose to the challenge and boiled down their insights in
each topic area to the key elements, practical consid-
erations, potential pitfalls, and helpful hints that this
text was designed to communicate. Chapters include
a few references for those seeking a more in-depth
treatment of the topic area.
Our hope in assembling this cogent overview of
the broad realm of clinical research is to provide a text
that both established and apprentice clinical research-
ers will find to be a useful guide and reference.
Flora M. Hammond, MD
James F. Malec, PhD
Todd G. Nick, PhD
Ralph M. Buschbacher, MD

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd xiv 8/11/2014 10:59:17 AM

 Share
Handbook for Clinical Research: Design, Statistics, and Implementation

Hammond_87543_PTR_00_FM_i-xiv_08-12-14.indd ii 8/11/2014 10:59:15 AM

 I
Design

Hammond_87543_PTR_01_1-8_08-12-14.indd 1 8/11/2014 10:59:32 AM

Hammond_87543_PTR_01_1-8_08-12-14.indd 2 8/11/2014 10:59:32 AM
 Eric Lenze, MD and John Whyte, MD, PhD
Development and Testing
of Treatments
DEFINITIONS AND DESCRIPTIONS
■ Active ingredients: The components of an interven-
tion under study that produce the therapeutic change
according to its underlying theoretical model.
■ Enablement/disablement theory: A class of theo-
ries about how changes in a proximal clinical target
(referred to as the target of treatment) will influence
distal clinical targets (referred to as treatment aims).
■ Target of treatment: The functional variable, that is,
hypothesized to be directly changed by the treatment
(eg, if leg strengthening exercises are performed in the
hope that ambulation will be improved, “leg strength”
is the target of treatment; improvement in ambulation
depends on many factors beyond the treatment, as
specified by enablement/disablement theory).
■ Treatment aim(s): Treatment goals that are distal
to the target(s) of treatment, and that are intended to
be indirectly improved, via mechanisms specified by
enablement/disablement theory (eg, ambulation in the
above example).
■ Treatment fidelity: The extent to which a study inter-
vention is being carried out as it was intended in the
protocol or manual (ie, that the specified active ingre-
dients are being delivered) and is clearly differentiable
from control intervention (eg, in a pilot or confirma-
tory randomized controlled trial [RCT]).
■ Feasibility: Demonstration that “moving parts” of
a clinical trial work (eg, recruitment, retention, mea-
surement reliability, and fidelity of the therapists to
the intervention).
Hammond_87543_PTR_01_1-8_08-12-14.indd 3
1
■ Proof of concept/proof of principle: Evidence sup-
porting the treatment theory (theory of the mecha-
nism of action by which the delivered “active
ingredients” of treatment produce change in an aspect
of functioning).
■ Minimum clinically important difference (MCID):
The minimal effect that has clinical relevance in the
management of patients.
INTRODUCTION
■ Imagine a researcher wishing to develop an entirely
new therapeutic intervention or treatment, or to adapt
an existing treatment for a new population. This chap-
ter describes the work needed prior to testing this new
treatment in an adequately powered RCT or other rig-
orous confirmatory design. This chapter distinguishes
treatment development research from treatment test-
ing research.
● Treatment development: The work needed
prior to doing a well-powered RCT or other rig-
orous confirmatory trial. Treatment development
includes:
■ Observational studies of the condition of
interest in the absence of specific treatment (or
in the context of failure to respond adequately
to an established treatment) may help to define
the state of affairs to be improved upon and to
identify the sites and time frames most suitable
for subsequent research.
8/11/2014 10:59:32 AM
 4 I: DESIGN
■ Theory development, open-label stud-
ies, small-pilot RCTs, and treatment manual
development.
■ All stages from the initial conceptualization
and protocolization of a new treatment, to the
pilot work done to demonstrate the feasibility of
a treatment study.
● Treatment testing refers to studies large enough
to be considered definitive tests of a treatment
hypothesis. Typically, these are well-powered
RCTs, often multicentered, but may include rigor-
ous quasi-experimental designs. Treatment testing
can be further divided into confirmatory testing of
the treatment theory and testing of the practical
clinical impact of the treatment.
■ In some cases, these goals can be combined—
when the target of treatment is also a clinically
relevant aim of treatment (eg, when strength-
ening exercises are performed so that heavier
items can be lifted, the target of “increased
strength” is a clinical goal in its own right).
■ In other cases, the best possible treatment of the
target may not lead to distal functional improve-
ments in all patients, as when a patient with leg
weakness and ataxia fails to improve in ambula-
tion (an aim) despite strengthening exercises that
improve the weakness (the target). In such cases,
one may perform a confirmatory trial demonstrat-
ing that the treatment reliably improves weak-
ness (the target), but one must assess its practical
impact on ambulation in a separate research with
a very different design (see below).
■ Confirmatory testing of the treatment theory
● Confirmatory tests of a treatment theory
should rigorously compare the effects of the
treatment versus some comparator on mea-
sured changes in the treatment target (effi-
cacy in treating the target).
● Later tests may explore whether those
treatment-induced changes in the target are
reliable across a heterogeneous population
of patients, and when delivered by hetero-
geneous clinicians (effectiveness in treating
the target).
■ Testing of the practical clinical impact of
treatment
● When testing for practical impact on a
distal clinical aim (eg, participation in social
or role activities), one must have an enable-
ment/disablement model in mind of how the
treatment target relates to the distal aim, and
what other strengths and limitations may
also affect the aim.
Hammond_87543_PTR_01_1-8_08-12-14.indd 4
● Enhancing treatment impact on a distal
clinical aim requires either:
1. Identifying patients for whom the target is
the primary limitation with respect to the
aim (eg, patients whose main obstacle to
ambulation is weakness); or
2. Developing a “package” of treatments
that address the multiple potential targets
that limit performance of the aim.
■ Research to flesh out the enablement/dis-
ablement model of a particular functional aim
area need not be carried out for each treatment
separately, that is, the cause of the change in
the target (ie, what treatment produced it) is not
relevant to how that change will affect distal
aims.
IMPLICATIONS
■ Treatment development is a critical step in getting to
evidence-based rehabilitative treatment.
■ Treatment development consists of three key steps:
● Step 1: Manualization or protocolization of the
intervention;
● Step 2: Conducting iterative case series, often
called an open-label trial, to test and modify both
the intervention and all aspects (such as inclusion
criteria, outcome measures) of the research;
● Step 3: Conducting a pilot RCT to demonstrate
feasibility of doing an RCT.
■ Treatment development is required prior to treat-
ment testing.
● One key goal of treatment development research
is to reduce the chances of failure (to demonstrate
efficacy, or effectiveness, of a promising treatment)
at the confirmatory RCT level.
■ A “negative result” is interpreted to mean
that the underlying treatment theory is false.
However, a negative result may also occur when
the treatment theory is correct but one of the
following failures occurred:
● Treatment is given in an inadequate dose.
● Treatment dose/duration is adequate to be
potent, but is poorly implemented by study
clinicians (ie, poor treatment fidelity).
● Treatment implementation by clinicians is
adequate, but patient adherence too low.
● Patient adherence to treatment is ade-
quate, but missing data rate is too high (due
to patient dropout/refusal or other reasons).
● Recruitment is poor, leading to inadequate
sample size.
8/11/2014 10:59:32 AM
 1: DEVELOPMENT AND TESTING OF TREATMENTS
● Outcome measures inadequately sensitive
to change.
● Outcome measures inadequately valid or
reliable.
● Treatment does not have the predicted
effect in the particular patient group cho-
sen (eg, due to too low, or too high severity/
impairment, comorbidities, and cognitive
issues).
● All of the above are adequate, but the con-
trol condition is inadequate (inadvertently
has active ingredients and is too potent, or is
not credible and leads to differential dropout
in a RCT).
■ Because there are so many opportunities for
failure at the RCT level, treatment development
for feasibility is designed to find these reasons
for failure and fix them, prior to moving to the
confirmatory RCT.
● In addition to the practical and feasibility issues,
treatment development research serves to provide
initial proof of concept (or lack thereof) for a new
treatment theory, providing an early “go or no-go”
decision on a treatment theory.
■ A proof-of-concept test is not the same as a test
of efficacy (or effectiveness), and should not be
expected to demonstrate the above components
at a rigorous level of certainty (eg, alpha = .05).
■ Proof-of-concept testing is different than effi-
cacy testing.
● Efficacy testing is done in a confirmatory
RCT that requires a priori calculations of
power and sample size, and predetermined
inclusion/exclusion criteria, outcome assess-
ments, intervention parameters, and analytic
strategy.
■ Proof-of-concept testing may return a nega-
tive result because the treatment theory is
incorrect or the treatment has not been made
sufficiently potent.
STRATEGIES
Treatment Development
■ Step 1: Treatment manual development
● Multidisciplinary team with expertise in treat-
ment development
■ Identify the “active ingredients”
■ Create a treatment manual that specifies active
ingredients
● The manual needs to accomplish two important
things
Hammond_87543_PTR_01_1-8_08-12-14.indd 5
5
■ Implement the underlying treatment theory
● For example, if the researcher is testing
a theory regarding treatment of working
memory, the manual should address work-
ing memory and the specific “ingredients”
required to modify it.
■ Be practical for use by the clinicians who are
doing the intervention in the study. Clinicians
who provide the treatment need to learn how to
provide the treatment consistently and do not
necessarily need an extensive description of the
underlying theory.
● Cannot manualize every aspect of a treatment
■ Decide what are the “essential ingredients” or
mechanisms of an intervention, according to the
theory that is being promoted.
■ Some manuals are highly protocolized, while
others are much less so.
● Typically in the evolution of a new treatment
intervention, the manual starts out brief, then
gets more extensive and more protocolized
(often with case examples) as the researchers
get more experienced with the active ingre-
dients, and then finally gets shorter and less
protocolized in efforts to make the treatment
implementable in the real world.
■ Manuals range from descriptions of a treat-
ment “approach” to step-by-step “how to” man-
uals, with tradeoffs.
● The former allows more flexibility in tai-
loring to individual patients, but provides
less specific guidance about exactly what the
clinician should do in specific situations.
■ Especially in the early phases of treat-
ment development, a manual is an evolving
document.
● Review and make changes based on advice
and information from collaborators:
– The study clinicians who are imple-
menting the intervention
– Debriefing patients involved in early
treatment delivery
● Query these individuals about what is
feasible:
– How much supervision of implementers
is needed, and how will it be done?
– What is likely to be well/poorly
received by the patients?
■ Feedback from patients, their families, and
providers is very helpful at this stage of treat-
ment development, and increasingly is recom-
mended or even required in grant applications
(eg, Patient-Centered Outcomes Research
Institute—PCORI—applications).
8/11/2014 10:59:32 AM
 6 I: DESIGN
■ The treatment manual development step is
usually fairly brief but continued evolution of
the manual is likely in ensuing phases.
■ In particular, if early treatment testing sug-
gests limited potency, the treatment may need
to be revised, and with it the manual.
■ Step 2: Conduct iterative case series
● The iterative cases series is similar to “Phase 1”
in drug treatment (Phase 1, Phase 2, and Phase 3)
parlance.
● Probably the most important step because it is
the main opportunity to gain experience with the
intervention and with the research methods, and to
make changes.
● In treatments for rapidly evolving conditions (eg,
during rapid natural recovery from a neurologic
event), case series, although useful for explor-
ing feasibility issues, may provide little evidence
regarding proof of principle, requiring introduction
of more rigorous control conditions earlier in the
treatment development process.
● Often this phase starts with wide inclusion and
(few) exclusion criteria.
■ Fine-tune these based on experience in
this phase; for example, the observation that
patients within a certain subgroup (defined by
demographics or clinical presentation) are par-
ticularly good, or particularly poor, subjects for
this intervention.
● Treat all participants (no control group), unless
a control group is required to assess proof of
concept.
● Keep tweaking intervention based on observa-
tions of where it appears to be inadequate.
● Gather feasibility and validity data on all
measures.
● Supervise study clinicians implementing the
treatment.
■ Often, and particularly early in this stage, the
research team will need to directly observe all
of the treatment sessions.
■ Provide feedback to implementing clinicians
regarding what is being done well and what
needs to change in order to achieve good treat-
ment fidelity.
● Measure fidelity.
■ It is often advisable to verify that the control
treatment is lacking in the treatment’s active
ingredients, even prior to a pilot RCT stage.
■ Fidelity assessment should focus on the
essential ingredients of the theory underlying
the new intervention.
■ In research treatment development and test-
ing, the following treatment fidelity steps should
Hammond_87543_PTR_01_1-8_08-12-14.indd 6
be implemented, documented, and detailed in
publications of the research:
● Treatment manual that defines treatment
● Details of training of therapists in the
treatment
● Details of supervision of therapists
● Fidelity monitored by external referees
● Fidelity data reported
● High level of fidelity attained and main-
tained in RCT
● Measure processes (ie, observable mechanisms)
when possible.
● The case series step is complete once the
researchers can see, consistently and to their satis-
faction, that the research methodology is feasible,
and that the intervention seems to be working:
■ Implementing clinicians carry out the inter-
vention with high fidelity.
■ Research methods for recruitment, retention,
measurement, and adherence are adequate.
■ Participants have a good treatment outcome
(and outcome measures can detect this).
● The main limitation of a case series is that par-
ticipants may appear to do well because of natu-
ralistic change, expectancy, or measurement error.
Thus, it is quite possible to gain a false sense of
confidence regarding feasibility and proof of con-
cept during a case series, which is then “lost” dur-
ing a pilot RCT (because the control group shows
similar change).
■ Step 3: Pilot RCT
● Pilot RCT is a randomized comparison of exper-
imental intervention to a comparator.
● Usually a necessary step in treatment develop-
ment, if the ensuing step is a confirmatory (well-
powered) RCT. However, many of the principles
below hold for other (quasi-experimental) designs.
● Uses measures of change in the treatment tar-
get as the primary criterion for assessing proof of
principle.
● The goals of a pilot RCT are:
■ To confirm feasibility as demonstrated by:
● Good recruitment and retention
● Even though recruitment and retention
for treatment may have been adequate in
the case series, introduction of blinding or
randomization may lead to a drop in recruit-
ment or retention that must be anticipated
and corrected.
● Comparator treatment is satisfactory (eg,
no excessive dropout, includes no active
treatment elements)
● Outcome and process measures are practi-
cal, reliable, and responsive
8/11/2014 10:59:33 AM
 1: DEVELOPMENT AND TESTING OF TREATMENTS
■ Satisfactory treatment fidelity:
● High adherence and competence in the
active group
● Effect of active treatment on treatment
target is clearly distinct from the control
intervention
● Demonstrate proof of concept as described
above but not to confirm efficacy or effectiveness
of the intervention.
■ A successful pilot RCT shows readiness for a con-
firmatory RCT.
● It is generally not expected to show a statisti-
cally significant effect, and it should not be used
either to report that a new intervention has been
“shown to be effective.”
● Additionally, the effect size in a pilot RCT may
not provide an accurate estimate for powering
future confirmatory studies (because the confi-
dence interval of the effect size in a small study is
often large, and because effect sizes in pilot RCTs
tend to be overly optimistic).
■ Instead, the MCID should be considered in
determining sample size in future studies (see
Chuang-Stein et al, 2011).
■ The appropriate method for determining the
MCID, however, remains controversial (see
Revicki et al, 2008).
● The pilot RCT, as well as the case series, are
publishable findings.
■ However, they should be reported not as efficacy
tests but rather as preliminary, feasibility, treat-
ment development, or proof-of-concept studies.
● The vast majority of intervention studies in med-
icine are pilot studies (but are often not identified
as such in publications).
■ The pilot RCT stage is also the last chance to adjust
various components of the intervention or other
aspects of the research methods (such as inclusion/
exclusion criteria or analytic strategy), as confirma-
tory RCTs usually “lock in” all design decisions at the
beginning of the study.
■ Although it is not always necessary to do a pilot
RCT prior to a confirmatory RCT, a pilot RCT is nec-
essary when:
● Recruitment (and retention) in randomized treat-
ments is uncertain.
● Feasibility/credibility of comparator condition
in the study population is unknown.
● The study team (or the research community) has
little research experience with either the experi-
mental intervention or with randomized tests of it.
● In a competitive funding environment, presenting
pilot data demonstrating feasibility and a possible
treatment effect may enhance the chances of funding.
Hammond_87543_PTR_01_1-8_08-12-14.indd 7
7
Confirmatory Testing of the Treatment
Theory (Efficacy Testing)
■ Design protocol for RCT or appropriate rigorous
study design
● Determine necessary sample size for adequate
statistical power to test both positive and negative
results with respect to the primary outcome(s).
Determine statistical analysis that will be used for
the main confirmatory analysis.
● Identify any a priori hypotheses about subgroup
differences in response to treatment, since only
pre-specified subgroup analyses are likely to be
trusted as confirmed findings, rather than hypoth-
eses for future exploration.
● Develop standardized approaches to training of
treaters, assessment of fidelity, recruitment and
retention, and data quality.
● Conduct regular surveillance of enrollment,
fidelity, and data quality throughout the trial.
● Conduct the pre-specified analyses when the
trial is complete.
Testing the Practical Impact of Treatment
(Effectiveness Testing)
■ Once a treatment has been confirmed to produce
robust impact on the target of treatment, consider
enablement/disablement models of the functional
domain to understand the relationship between the
treated target and other related areas of real world
function. What kinds of distal impacts are plausible
given these models?
■ Design subsequent studies that:
● Explore distal impact in patients with impair-
ment limited to the treated target (who should have
distal benefit).
● Explore distal impact in patients with multiple
impairments who receive a “package” of treat-
ments that address multiple relevant targets.
PITFALLS
■ A new treatment may potentially fail or be sub-
optimal at any of these steps. Although this some-
times requires abandoning the treatment theory, in
many cases it requires going back a step or two
in the treatment development stages, and repeating
steps.
■ Step 2 (iterative case series) is often the most
important part of treatment development but also
the hardest to get buy-in from grant or journal
8/11/2014 10:59:33 AM
 8 I: DESIGN
reviewers (who prefer to see an RCT design). In
grant applications, it is important to clearly explain
the critical role of the proposed case series in
the planned stages of the specific new treatment
development.
■ Treatment fidelity measurement is nonnative to
some clinical fields, and attaining and demonstrat-
ing good treatment fidelity may be a novel skill set
for researchers. See Hildebrand et al (2012) for an
empiric example of this process.
■ See Whyte and Barrett (2012), for a more detailed
description of the stages of rehabilitation treatment
development and their pitfalls.
HELPFUL HINTS
■ Each of the various treatment development phases is
essential. Do not skip or speed through steps.
■ Get help from experts in treatment research.
■ Ensure adequacy of measurement (completion, reli-
ability/validity, and sensitivity to change) before mov-
ing to RCT phases.
Hammond_87543_PTR_01_1-8_08-12-14.indd 8
SUGGESTED READINGS
Chuang-Stein C, Kirby S, Hirsch I, Atkinson G. The role of
the minimum clinically important difference and its impact
on designing a trial. Pharm Stat. 2011;10(3):250–256.
Hildebrand MW, Host HH, Binder EF, et al. Measuring
treatment fidelity in a rehabilitation intervention study.
Am J Phys Med Rehabil. 2012;91(8):715–724.
Revicki D, Hays RD, Cella D, Sloan J. Recommended meth-
ods for determining responsiveness and minimally impor-
tant differences for patient-reported outcomes. J Clin
Epidemiol. 2008;61(2):102–109.
Thabane L, Ma J, Chu R, et al. A tutorial on pilot studies: the
what, why and how. BMC Med Res Methodol. 2010;10:1.
Retrieved from http://biomedcentral.com/1471/2288/10/1.
Whyte J. Contributions of treatment theory and enablement
theory to rehabilitation research and practice. Arch Phys
Med Rehabil. 2014;95(Suppl 1):S17–S23.
Whyte J, Barrett AM. Advancing the evidence base of reha-
bilitation treatments: a developmental approach. Arch
Phys Med Rehabil. 2012;93(8 Suppl):S101–S110.
Whyte J, Hart T. It’s more than a black box; it’s a Russian
doll: defining rehabilitation treatments. Am J Phys Med
Rehabil. 2003;82(8):639–652.
8/11/2014 10:59:33 AM
 Elena Gillespie, PhD
Qualitative Research
DEFINITIONS AND DESCRIPTIONS
■ Epistemology: The categories of knowledge that a
culture and society consider of value.
● Example: A scientist of Medieval Europe would
recognize astrology as a valid science, whereas a
modern scientist, in general, would not do so. The
body of information, practice, and application of
astrology would not have changed over that same
time period.
■ Phenomenology: Observing and analyzing data, the
research participant and their environs for the subjec-
tive experience of the participant.
■ Hermeneutics: Reflexively analyzing data situ-
ated within investigator’s biases and epistemological
values.
● Example: A patient suffers a severe brain
injury; in observing the patient’s family interac-
tions and responses, it becomes apparent that
the family believes that the patient will experi-
ence a full recovery. After a series of interviews
with the practitioner, the practitioner understands
that the family members are not intellectually
challenged in their ability to understand what is
being explained to them, but are not familiar with
the medical terminology used and the course of
recovery. The practitioner also understands that
the family is undergoing a period of emotional
stress, so that extra time and care must be taken
to explain possible outcomes, perhaps multiple
times, in a manner, that is sensitive to their social
and cultural milieu. The investigator has followed
a hermeneutic process based on their experience
and observations.
Hammond_87543_PTR_02_9-12_08-12-14.indd 9
2
■ Reflexivity: To repeatedly examine the interrelation-
ships within a set of data (eg, interviews) in a manner
that appreciates the investigator’s impact on the data.
● Investigators can and often do impact the qual-
ity of the data gathered; it is their responsibility
to understand how their beliefs influence their
observations.
■ Bracketing: To set aside an observation within a set
of data without assessing for meaning to examine other
data within the set for the significance of those data.
● For example, correlation of two variables may
not equal causation, although it might. Each vari-
able is isolated and analyzed before such a conclu-
sion is reached (if it can be—there may be other
variables present influencing the outcome).
■ Ethnography: Field studies or case reports exploring
cultural or social phenomena.
● Example: The examination of the social order
and hierarchy in Filipino families and their effects
on the self-image of individuals with spinal cord
injury.
■ Ethnomethodology: The study of common socially
determined methods and rituals for producing and
maintaining social order.
● Example: The examination of the decision-
making process utilized by American families
precipitant to installing an elderly family member
into a nursing home. These would include factors,
such as the family dynamic, financial outlook, and
health needs of the family member in question.
■ Grounded theory: Post hoc exploration of the data to
identify observations and trends not identified a priori
to the study. The analysis of data is led or grounded by
where the data leads; the investigators return repeatedly
8/11/2014 10:59:47 AM
 10
to assess the data for the next step. The hypothesis is
then formulated based on where the information led
the investigators.
■ Field study: A set of data—usually interviews—
gathered from participants in the area of interest.
● Involves narratology, that is, storytelling
● Can be structured, semi-structured, or nonstruc-
tured interviews
■ Participatory action research (PAR): Research in
which the participants not only provide data but are
also asked about their experience(s) with the topic and
are actively enlisted to create the study’s hypothesis.
● The investigators concern themselves with
the methodology and implementation. PAR may
involve specifically designing a study to increase
the understanding of the participants regarding a
particular social or medical issue and involve them
in social change related to that issue.
● Example: Several groups of the relatives of
patients with traumatic brain injury (TBI) are
enlisted to discuss how best to obtain access to
federal monies to aid the patients to return to work
more effectively, with guidance from the investiga-
tors. The study then continues to monitor the group
and determines how many of them become activ-
ists as a result of their new level of understanding
of federal processes and TBI.
■ Case study: A write-up of an individual case to
exemplify either its singular qualities different from
the mean of the illness or syndrome, or to emphasize
typifications across the population. Can be descrip-
tive, analytic, or collective (or more than one).
■ Critical social research: Analyzing the means of
communication within interviews to determine how
subjects derive meaning.
● Example: A field study of a group of patients to
assess how their sense of identity has changed after
a stroke.
■ Ethical research: Analysis of the foundations of an
area to determine its ethical issues and problems.
● Example: Our society’s adherence to consis-
tently constructing barrier-free access for individu-
als with disabilities in the United States is believed
to be erratic. To determine factors contributing to
this inconsistent implementation, a field study is
undertaken of subjects who have returned to work
after a spinal cord injury documenting the physical
obstacles they must navigate and assistive devices
they use every day to accomplish this goal. The laws
and regulations that govern their implementation are
also investigated for their consistency of application
and congruence to real-life obstacles to determine,
from an ethical perspective, should these laws or
access to these assistive devices be changed.
Hammond_87543_PTR_02_9-12_08-12-14.indd 10
I: DESIGN
■ Mixed methodology: A combination of both quan-
titative and qualitative methods in the conduct of a
clinical research study.
● Example: A study of the frequency, locations,
and size of injection doses of Botulinum toxin A
for the amelioration of spasticity in patients with
stroke, and interviewing a subset of these subjects
in a structured interview about the quality of their
mobility over a 3-month period.
■ Content analysis: Themes in the transcribed inter-
views are identified and coded, then further broken
down into more minute similarities and differences,
and organized into categories.
INTRODUCTION
■ Qualitative methodology is the study of the individ-
ual subject’s perceived experience.
■ Assessing the “mean” or average experience of
many is not its goal, but rather the purpose of qualita-
tive research is the identification of a few variables
that impact a small sample, that would probably not
come to light in a quantitative study.
● Rigor—clarity and the economical application
of logic—are fervently applied in its utilization.
● Central criteria: Can the answer to the question add
knowledge to the field of inquiry about this topic?
IMPLICATIONS
■ Can deliver deep insights into treatment effects and
clinical impact not measurable by quantitative surveys
or other measures.
■ Not appropriate for use for quantitative measure-
ment, that is, provides only nominal level data.
BACKGROUND
■ Crisis of representation arose in anthropology stud-
ies in the early 1900s.
● Is what is observed and documented in cultural
and social studies truly representative of the group
or an artificial construct of the group imposed by
the investigators, or altered by the subjects, know-
ing they are being observed?
■ Critical theory of Adorno and Horkheimer of the
Frankfurt School of Sociology of the ‘30s and ‘40s
reflexively examined Positivism and the Scientific
Method and postulated that significant meaning was
being lost for the research participant as an individual
in the resulting reductive mindset.
8/11/2014 10:59:47 AM
 2: QUALITATIVE RESEARCH
■ Edmund Husserl, Martin Heidegger, and Hans-
George Gadamer added to the area of inquiry by
assisting to define ontology—the integration of human
consciousness and its perceptions within the real
world. Gadamer in particular described the “fusion of
horizons” of understanding through language (herme-
neutic approach).
■ Foucault observed that epistemology—what we
value as knowledge and define as “data”—is driven
by the predominant group in power, not by “truth.”
■ Max van Manen, Clark Moustakas elaborate on
writing and storytelling as a means of exploration.
STRENGTHS
■ Can consider questions or clinical observations that
are not captured by quantitative measures and that
impact clinical outcomes.
■ Qualitative measurement is a highly useful addition
to a clinical study in which the primary end point is a
quantitative measure.
● Example: A placebo-controlled, double-blind
randomized controlled trial (RCT) for efficacy of
an intervention also conducts a blinded field study
of a subset of subjects from each group and discov-
ers that the practitioners’ clinical manner and style
in each arm of the study were widely different, thus
impacting how the intervention’s effectiveness was
perceived, and consequently, the strength of the
placebo effect.
WEAKNESSES
■ Transcription is time-consuming and laborious.
■ Methodology is not widely accepted by funding
agencies or journals.
STRATEGIES
■ Reliability can be addressed by
● Multiple observers
● Interviewer corroboration
● Research participant check (participation checks
transcript for errors or misunderstandings)
● Confirmation with larger studies
■ The investigator must know him or herself well;
that is, his or her biases and background, and docu-
ment them before beginning. The investigator is the
“metacode,” that is, the medium by which the data are
filtered and interpreted.
Hammond_87543_PTR_02_9-12_08-12-14.indd 11
11
● Example: A qualitative investigator does not
believe that socioeconomic class impacts treat-
ment compliance and therefore unconsciously only
recruits subjects that own their own car for trans-
port to the study site.
■ The investigator is a part of the study. Meticulous
field notes and observations are kept on the investiga-
tor’s experience in relation to participant enrollment
and data collection.
● Example: In a group round table setting, the
investigator, after listening to the audio recording
multiple times, finds that he or she subconsciously
interrupted subjects who were reflecting an oppos-
ing perspective to the investigator’s hypothesis,
thus effectively discounting that subject’s data and
impacting the flow of the discussion.
■ Requires Institutional Review Board (IRB) approval
and informed consent document. Using patient names,
demographics, and other significant identifiers is not
permissible in the final manuscript.
■ Recruitment similar to RCTs; advertisements, word-
of-mouth, and letter campaigns.
■ Sample is generally no greater than nine
participants.
■ Inclusion and exclusion criteria are designated
before study initiation. These may change—document
the rationale for such change.
PITFALLS
■ Results represent subjects’ personal subjective expe-
rience; care must be taken not to extrapolate data to an
unsampled group.
● Example: A field study of the attitudes of
Hispanic men toward paraplegia after a disabling
accident could not be directly correlated to Hispanic
women, due to cultural and gender differences.
■ Qualitative studies are not designed to find the “aver-
age”; therefore, a large sample may lead to “investiga-
tor drift,” that is, desensitization to the information
gathered.
● Small details can be missed in the face of a large
amount of transcription and analysis.
■ If interviews are chosen as the means of data col-
lection, think through the structure of the interview
carefully, that is, whether to leave it as structured,
semistructured, or unstructured.
HELPFUL HINTS
■ Do not overenroll. Six to nine subjects will provide
a rich continuum of information to examine.
8/11/2014 10:59:47 AM
 12
■ Return to older notes repeatedly to continually
assess whether identification of the central themes has
changed due to the actual data, or from investigator
bias and “burn-out.”
■ Organization and attention to detail are key.
■ There are a number of different software applica-
tions that can assist the investigator (see Resources for
recommendations).
● Categories can be statistically measured if nec-
essary, but the sample is usually small, so statisti-
cal analysis is not applicable or revelatory.
● A qualitative data management program is
only as good as the investigator; it is really only
a “super-organizer.” Original thought, rigor, and
discernment still has to come from the investiga-
tor. One can accomplish the same goal by using a
corkboard, colored paper, and stickpins.
SUGGESTED READINGS
Denzin NK, Lincoln YS. The Sage Handbook of Qualitative
Research. Thousand Oaks, CA: Sage Publications; 2005.
Hammond_87543_PTR_02_9-12_08-12-14.indd 12
I: DESIGN
Moustakas C. Phenomenological Research Methods.
Thousand Oaks, CA: Sage Publications; 1994.
Palmer RE. Hermeneutics. Chicago, IL: Northwestern
University Press; 1969.
Pope C, Mays N. Qualitative Research in Health Care.
Malden MA: BMJ Books; 1996.
van Manen M. Researching Lived Experience: Human
Science for an Action Sensitive Pedagogy. London,
Ontario Canada: State University of New York; 1990.
RESOURCES
Software
■ Open source:
● Coding Analysis Toolkit (http://cat.ucsur.pitt.edu/)
● RQDA (http://rqda.r-forge.r-project.org/)
● QDA Miner Lite (http://provalisresearch.com/
products/qualitative-data-analysis-software/freeware/)
● Aquad (http://www.aquad.de/en/)
■ Proprietary:
● Atlas.ti (http://www.atlasti.com/index.html)
● HyperRESEARCH (http://www.researchware.com/
products/hyperresearch.html)
8/11/2014 10:59:47 AM
 Robyn L. Tate, MPsychol, PhD
and Michael Perdices, MA (ClinNeuropsychology), PhD
Single-Case
Experimental Designs
DEFINITIONS AND DESCRIPTIONS
■ The single-case experimental design (SCED) is the
prospective and intensive study of a single individual
who serves as his or her own control.
■ An SCED uses experimental methodology and
therefore does not include the case description or case
report, even if quantitative data are provided.
■ Essential features of the SCED:
● Contains multiple, discrete phases (or time periods)
● Phases are of two types:
■ Baseline/no treatment phase, usually desig-
nated as A
■ Treatment/intervention phase, usually desig-
nated as B
■ Other letters (eg, B , C, D) are used to indi-
1
cate different treatments.
● The independent variable(s) (ie, treatment
or intervention) is systematically manipulated
(applied and withdrawn) across the phases.
● The dependent variable (ie, symptom being
treated or the target behavior) is measured repeat-
edly and frequently throughout each phase.
■ Other terms used for SCEDs:
● Medical sciences use the term “n-of-1 trial,”
specifically referring to the randomized, multiple
crossover design in a single patient.
● Behavioral sciences also use terms such as “sin-
gle-case design,” “single-participant design,” and
“single system design” (the older literature also
uses the term “single-subject design”).
Hammond_87543_PTR_03_13-17_08-12-14.indd 13
3
INTRODUCTION
■ SCED is useful when
● It is not feasible to conduct a randomized con-
trolled trial (RCT), such as infrequently occurring
conditions
● The results of an RCT may have limited applica-
tion for individuals whose personal and/or clinical
characteristics differ substantially from selection
criteria used in RCTs (eg, excluding patients with
comorbidities).
■ In these situations, the SCED provides a method
to empirically test the effect of an intervention on a
single participant.
■ Well-designed SCEDs can demonstrate a causal
relationship between the treatment and the symptom
being treated or target behavior.
■ Many types of SCEDs are available, which are appro-
priate for different types of interventions and suitable
for answering different types of research questions.
■ This chapter describes the three most common types
of SCEDs and their strengths, weaknesses, and chal-
lenges compared to the traditional group-based design.
IMPLICATIONS
■ In 2011, the Oxford Centre for Evidence-Based
Medicine (http://www.cebm.net/index.aspx?o=5653)
introduced the Levels of Evidence 2 table in which
the n-of-1 trial is classified as Level 1 evidence for
8/11/2014 11:00:05 AM
 14
treatment decision purposes in the individual patient,
comparable to the systematic review of multiple
RCTs.
■ SCEDs can serve as a good prelude to a Phase I
clinical trial.
■ SCEDs may be the only means of empirically estab-
lishing the effectiveness of a treatment in an individ-
ual patient.
■ SCEDs provide an opportunity to use evidence-
based principles in everyday clinical practice.
■ SCEDs encourage involvement from the partici-
pant in treatment decision-making processes, thereby
increasing the social value of the intervention.
BACKGROUND
■ The main types of SCEDs include
● Withdrawal/reversal designs (also called cross-
over designs; see Figure 3.1):
■ Generally consist of a series of baseline (A)
and treatment (B) phases (eg, A-B-A-B; A-B-
A-C), which may be either counterbalanced
(eg, A-B-A-B-A-B) or randomly sequenced
(eg, A-B-B-A-B-A);
■ Are used when the treatment can be meaning-
fully withdrawn (eg, medications, equipment,
technical aids).
● Multiple baseline designs (see Figure 3.2):
■ The dependent variable is measured across a
number of individuals, settings, or behaviors.
■ “Multiple baseline”:
● Denotes the introduction of the same
treatment in a staggered sequence across
individuals, settings, or behaviors
● Does not refer to obtaining multiple mea-
sures of the dependent variable during the
baseline phase
4
Target Behavior
1
A Baseline
0
1 3 5 7
FIGURE 3.1 Hypothetical data for a withdrawal (A-B-A-B) design.
Hammond_87543_PTR_03_13-17_08-12-14.indd 14
I: DESIGN
■ Are used when there are carry-over effects
and treatment cannot be meaningfully with-
drawn (eg, acquisition of skills, therapeutic
instructions).
● Alternating treatment designs (see Figure 3.3):
■ Two or more treatments are compared con-
currently in an alternating manner, rather than
sequentially as in the withdrawal design.
■ A baseline condition is not necessary.
STRENGTHS
■ SCEDs that are methodologically rigorous (see
Strategies) and have the capacity to determine
cause–effect relationships between the dependent
and independent variables, thereby establishing
empirically validated treatments and evidence-
based practices.
■ SCEDs can determine whether the treatment has
been effective for an individual patient, whereas with
an RCT, information on the number of participants
who improve and their characteristics is not (or only
rarely) provided.
■ SCEDs are less costly to conduct than large or mul-
ticenter RCTs.
■ SCEDs can provide a means to fractionate and iso-
late components of combination treatments to deter-
mine which components are effective.
■ SCEDs can be used with heterogeneous populations
and rare conditions.
■ SCEDs have the flexibility to tailor treatment to an
individual’s specific needs and changing conditions.
■ SCEDs are suitable when there are ethical objec-
tives to withholding treatment for a prolonged period
of time (eg, if randomized to the control group).
■ SCEDs provide a good methodological framework
for clinicians to conduct research.
B Treatment A Baseline B Treatment
1 3 5 7 9 11 13 15 17 19 21 23
Sessions
8/11/2014 11:00:05 AM
 3: SINGLE-CASE EXPERIMENTAL DESIGNS 15

Target Behavior 1 20

15

10

5
A Baseline B Treatment
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Sessions

20
Target Behavior 2

15

10

5
A Baseline B Treatment
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Sessions

20
Target Behavior 3

15

10

5
A Baseline B Treatment
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Sessions

FIGURE 3.2 Hypothetical data for a multiple baseline design across behaviors.

20
18
16
Target Behavior

14 Treatment 1
12
10 Treatment 2
8
6
4
2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Session

FIGURE 3.3 Hypothetical data for an alternating treatment design.

Hammond_87543_PTR_03_13-17_08-12-14.indd 15 8/11/2014 11:00:05 AM

 16
WEAKNESSES
■ Poorly designed SCEDs are subject to risk of bias
and threats to internal validity (see Strategies that, if
applied, will strengthen internal validity).
■ The main threat to external validity is generaliza-
tion (ie, will the treatment be effective for other
individuals?).
● This threat is addressed by replication across
individuals.
● Generalization is not an issue in the clinical set-
ting when the question is whether the treatment
was effective for the specific patient being treated.
■ It is important to establish baseline stability, but
there are no specific or agreed-upon criteria for its
determination.
■ SCEDs need commitment on the part of the
researcher and participant because generally, assess-
ment and therapy sessions are frequent and can be
labor intensive.
CONTROLS
■ In an SCED, the participant serves as his or her own
control, by having the investigator manipulate the
phase (Condition 1, Condition 2).
STRATEGIES
Before the Study Begins
■ Conduct a behavioral analysis to:
● Identify the symptom/target behavior to be
treated
● Identify factors that are maintaining the target
behavior
● Define and operationalize the symptom/behav-
ior in precise, specific, and objectively quantifiable
terms
● Determine how the symptom/behavior is to be
measured, ensuring that the measures are suitable
for repeated administration (and hence need to be
brief)
■ Suitable measures may include
● Frequency counts (eg, number of times a
symptom/behavior occurs)
● Rating scales or checklists (eg, for pain,
depression)
● Correct performance scores
● Source any stimulus materials required for mea-
surement of the symptom/behavior
Hammond_87543_PTR_03_13-17_08-12-14.indd 16
I: DESIGN
■ Identify the design that is best suited for the inter-
vention and the clinical research situation (see
Background).
■ Consider the use of randomization. In SCEDs, ran-
domization refers to:
● The onset of each phase
● The order of the phases (crossover designs)
● The order of the treatment pairs (alternating
treatment designs)
■ Ensure that the assessor is independent of the thera-
pist, as well as blind to the phase of treatment.
■ When possible, blind the participant and the person
conducting the intervention to the phase of treatment
(eg, baseline vs treatment).
■ In pharmacological treatments, plan for run-in and
washout periods.
■ Obtain institutional ethics approval if the SCED is
being used as a research tool.
■ Involve the patient and family in selecting the symp-
tom/target behavior for treatment, its measurement,
and the intervention.
After the Study Starts
■ Establish interrater reliability of the measure for
each phase (recommended for ≥20% of observations
per phase, with recommended result of ≥80% agree-
ment/kappa ≥0.6; see Kratochwill et al, 2010).
■ Ensure sufficient sampling of behavior occurs in
each phase (recommended ≥5 data points per phase;
see Kratochwill et al, 2010).
■ Plot every data point obtained from each phase onto
a graph.
■ Analyze data using appropriate techniques (see
Perdices and Tate, 2009, for overview):
● Traditional visual analysis, which includes sys-
tematic evaluation (see Kratochwill et al, 2010) of
between-phase changes in level, and/or slope, and/
or variability
● Visual analysis accompanied by quasistatistical
techniques (eg, celeration line, 2 standard devia-
tion band)
● Statistical analysis:
■ Parametric (eg, t test, F test), for ratio/interval-
level data and when there is no autocorrelation
■ Nonparametric (eg, Wilcoxon signed ranks
test), for ordinal-level data or when there is
autocorrelation in ratio/interval-level data
■ Time-series analysis (eg, C-statistic; Auto-
Regressive Integrated Moving Average
[ARIMA])
■ Percentage of nonoverlapping data
8/11/2014 11:00:06 AM
 3: SINGLE-CASE EXPERIMENTAL DESIGNS
PITFALLS
■ Ethical issues can arise if there is a conflict between
establishing a sufficiently long baseline to ensure sta-
bility and the need to initiate treatment promptly, if
patient safety is an issue.
■ In behavioral therapies it is almost impossible to
blind the participant and therapist.
HELPFUL HINTS
■ With pharmacological and complementary and
alternative medicine interventions, blinding of both
researcher and participant is possible, thus strengthen-
ing the internal validity of the investigation.
■ Consult with your institutional ethics board regard-
ing requirements for approval and regulation.
SUGGESTED READINGS
Barlow DH, Nock MK, Hersen M. Single Case Experimental
Designs. Strategies for Studying Behaviour Change. 3rd
ed. Boston, MA: Pearson; 2009.
Kratochwill TR, Hitchcock J, Horner RH, et al. Single-case
designs technical documentation. 2010. Available at http://
ies.ed.gov/ncee/wwc/pdf/wwc_scd.pdf.
Hammond_87543_PTR_03_13-17_08-12-14.indd 17
17
Perdices M, Tate RL. Single-subject designs as a tool
for evidence-based clinical practice: are they unrec-
ognised and undervalued? Neuropsychol Rehabil.
2009;19:904–927.
Tate RL, Perdices M, Rosenkoetter U, Wakim D, Godbee K,
Togher L, McDonald S. Revision of a method quality rat-
ing scale for single-case experimental designs and n-of-1
trials: The 15-item Risk of Bias in N-of-1 Trials (RoBiNT)
Scale. Neuropsychol Rehabil. 2013; 23(5):619–638.
Tate RL, Togher L, Perdices M, McDonald S, Rosenkoetter
U, on behalf of the SCRIBE Steering Committee.
Developing reporting guidelines for single-case experi-
mental designs: the SCRIBE project (Abstract. Presented
at 8th annual conference of the Special Interest Group in
Neuropsychological Rehabilitation of the World Federation
for NeuroRehabilitation). Brain Impair. 2012;13:135.
Vohra S, Shamseer L, Sampson M, et al. for the CENT
group. CONSORT Extension for N-of-1 Trials (CENT)
2012 Statement. (in preparation).
RESOURCES
PsycBITE (http://www.psycbite.com) has archived more
than one thousand designs using single-case methodol-
ogy, which are also rated for method quality
Reporting guidelines, in the CONSORT tradition, for both
single-case experimental designs (see Tate et al., 2012) and
n-of-1 trials (see Vohra et al.) are currently being prepared.
8/11/2014 11:00:06 AM
 4 Michael Schönberger, PhD, Dipl.-Psych.
Studies of Associations
DEFINITIONS AND DESCRIPTIONS
■ Biological gradient/dose–response curve: A form of
the relationship between two variables; for example,
amount of medication administered plotted against
subjective well-being
■ Repeated measurement studies: Studies in which
participants are measured on more than one occasion
■ Cross-lagged panel designs: The relative predictive
power of several variables is compared in a repeated
measurement study. Provides evidence regarding the
direction of causality between two or more variables
■ Path analyses: A statistical technique in which a
model of interrelationships between variables that
have been defined by the investigator is tested using
empirical data
■ Structural equation modeling: A form of path analy-
sis that contains latent variables (ie, factors underlying
observed variables)
■ Multilevel modeling: A flexible form of multiple
regression that is appropriate when the data structure
is hierarchical (such as patients measured in several
units in several hospitals; repeated measurement of
participants)
■ Multicollinearity: Strong correlations between pre-
dictors in a regression analysis; makes it difficult to
determine each individual predictor’s contribution to
the prediction of the dependent variable
■ Explorative factor analysis (EFA): Statistical tech-
nique with which dimensions underlying a set of vari-
ables/test items are determined
■ Confirmatory factor analysis (CFA): Statistical tech-
nique with which a model of dimensions underlying
Hammond_87543_PTR_04_18-22_08-12-14.indd 18
a set of variables/test items is tested using empirical
data
■ Cluster analysis: Grouping of individuals into sev-
eral clusters on the basis of their characteristics as
measured by a set of variables
INTRODUCTION
■ Studies of association are concerned with relation-
ships between measured quantities.
● Such associations can have different forms (linear
or nonlinear) and they can be causal or noncausal.
● An example of an association that involves a
causal relationship is the association between
patients’ engagement in rehabilitation (cause) and
their outcome (effect), but even this simple exam-
ple demonstrates the complexity of cause–effect
relationships in clinical research: Patients’ experi-
ence of improvement might well have a positive
effect on their engagement in rehabilitation. Often,
complex causal models are required to describe
clinical reality appropriately.
■ Associations can exist between variables with all
kinds of scale properties—nominal, ordinal, interval,
and ratio scales.
● The appropriate statistical procedure to deter-
mine associations between two or more variables
depends on
■ The number of dependent and independent
variables in the analysis
■ The variables’ scale properties
■ The distribution of values on these scales
8/11/2014 11:00:19 AM
 4: STUDIES OF ASSOCIATIONS
■ The assumptions regarding interrelation-
ships between predictor variables and between
dependent variables
■ Most importantly, the research question that
one is aiming to answer
■ The purpose of this chapter is to give an overview
of research designs and statistical techniques used
for studies of association as well as examples of
application.
IMPLICATIONS
■ In clinical research, studies of association are often
used to predict patient outcomes and to investigate the
processes that lead to these outcomes (eg, patient and
therapist characteristics and the quality of the client–
practitioner relationship).
■ In clinical practice, service providers are required
to secure and improve the quality of their services.
Studies of association can help to determine which
services contribute to a positive outcome.
BACKGROUND
■ From a philosophy of science point of view, causal
relationships cannot be proven. Hill defined criteria
that should be considered when deciding whether a
causal relationship between two variables should be
assumed:
● Strength of association
● Consistency across persons, places, circum-
stances, and times
● Specificity of association (eg, specific clinical
services are associated with specific outcomes)
● Temporality of association (what came first?)
● Biological gradient (dose–response curve)
● Plausibility
● Coherence of empirical findings
● Change in the dependent variable as an experi-
mental condition is changed
● Analogy of empirical findings to established
causal relationships
■ In studies of association, the strength of the asso-
ciation should always be determined and expressed
both in terms of statistical effect size and clinical
relevance.
■ Repeated measurement studies, such as cross-
lagged panel designs, can support causal assumptions
because they allow it to determine which variables
measured at an earlier time point affect other variables
measured later on.
■ In recent years, the statistical methods used in stud-
ies of associations have become more sophisticated.
Hammond_87543_PTR_04_18-22_08-12-14.indd 19
19
● Path analyses and structural equation modeling
are often used to test assumptions about complex
relationships between variables.
● Multilevel modeling is appropriate for the analy-
sis of multicenter study data.
STRENGTHS
■ Because studies of association usually do not rely
on an experimental research design, existing data-
bases can be used. This opens the possibility for the
use of large samples and hence very good statistical
power.
■ Especially when prospective and repeated mea-
surement designs are used, studies of associations
allow detailed examination of the complex interplay
between factors related to outcome after clinical inter-
ventions (eg, by using path analysis and structural
equation modeling).
WEAKNESSES
■ In the case of convenience samples, representative-
ness may be questionable.
■ No experimental variation of conditions.
■ Evidence for causality is weaker than in randomized
controlled trials.
STRATEGIES
Sample
■ Often a single group of subjects
■ As is always the case in clinical research, ideally the
sample should be randomly selected from the target
population in order to secure representativeness of the
sample.
■ However, studies of association often use conve-
nience samples (such as hospital databases)
■ Studies of association can be single-site or multisite
(eg, multicenter study)
Design Options
■ Cross-sectional (associations between variables at a
single time point) versus repeated measurements over
time (eg, cross-lagged panel designs with repeated
measurement of two or more variables and determina-
tion of their temporal relationship; Figure 4.1)
8/11/2014 11:00:19 AM
 20 I: DESIGN

■ Retrospective (measuring/making use of data from
the past; eg, comparative effectiveness studies) versus
prospective (real-time, single cross-sectional mea-
surement or repeated measurements over time)
■ Single-site studies versus multicenter studies
Methods (See Specific Chapters in Part II
for More Information About Statistical
Methods)
■ If an association between two variables is to be
determined: Compute a correlation, t test, analysis
of covariance, or their nonparametric equivalents,
depending on the quality of the measurement scales
and distributions.
■ If the relationship between two variables is to be
determined and the effect of a third variable is to be
partialled out: Partial correlation.
■ If a continuous dependent variable is to be predicted
from one or several predictors: Standard multiple
regression.
● Careful with predictors that correlate strongly
with each other (multicollinearity problem).
■ If a dichotomous/ordinal variable is to be predicted from
one or several predictors: Logistic/ordinal regression.
■ If one or several dependent variables are to be pre-
dicted from one or several observed variables, and the
interrelationships among the predictors and among the
dependent variables are to be modeled: Path analysis,
for example,
● Cross-lagged panel analysis testing temporal
relationships. This technique provides evidence
regarding the direction of causality between two
or more variables. In the example (Figure 4.1),
functional limitations following traumatic brain
injury predict depression, but depression does not
predict functional status.
● Test of mediator relationships. In the example
(Figure 4.2), the difference between brain-injured
individuals and healthy controls on a complex
selective attention task is mediated by group differ-
ences on a simple selective attention task, but not
by group differences on a working memory task.
● Test of moderator relationships (multigroup
path analysis or addition of interaction terms to the
model): Using multigroup path analysis, whether
the same explanatory model of interrelationships
between variables holds true in different popula-
tions can be tested.
■ In order to determine whether the relationship
between two variables is moderated by a third
variable, an interaction term can be added to a
predictive model.
■ If a path analysis contains factors that are not
observed directly but measured by a set of indica-
tors (eg, depression and anxiety, measured with the
Hospital Anxiety and Depression Scale): Structural
Equation Modeling.
● The first step in Structural Equation Modeling
is to compute one or several confirmatory factor
analyses in order to test whether sets of items are
indicators of underlying factors.
● Once this is confirmed, then assumptions regard-
ing the interrelationships between these underlying
factors (and possibly also additional, observed
variables) can be tested.
● In the example in Figure 4.3, mood changes
as well as cognitive and behavioral changes were
measured with several questionnaire items.

GOSE .62 GOSE

Z1
6 months 12 months

.09
.27 –.32
–.31
SCID Depression SCID Depression
Z2
6 months .34 12 months

FIGURE 4.1 An example for a path analysis testing the temporal association between functional status (GOSE) and
depression after traumatic brain injury in the form of a cross-lagged panel design analysis.
Note: GOSE = Glasgow Outcome Scale-Extended (measure of functional status); SCID = Structured Clinical Interview for DSM-
IV-TR. Curved lines represent correlations; straight lines represent regression coefficients. Correlations and standardized regression
coefficients are displayed. Z1 and Z2 are the residuals of the dependent variables. Coefficients in the model with P < .05 are printed
boldly. The analysis was statistically controlled for covariates age, gender, level of education, and PTA duration. These covariates are
omitted from the figure.
Source: Originally published in Schönberger et al (2011a). Reprinted with the publisher’s permission.

Hammond_87543_PTR_04_18-22_08-12-14.indd 20 8/11/2014 11:00:19 AM

 4: STUDIES OF ASSOCIATIONS 21

Z1 Group Z2

1 4 – 1

LNS SSAT

– 8

CSAT

Z3

FIGURE 4.2 Example of a path analysis testing a mediator relationship: Can the difference between head-injured
and nonhead-injured individuals on a complex selective attention task be explained by (is it mediated by) their per-
formance on tests of working memory and simple selective attention?
Note: Group = Group membership (head injury vs healthy control); LNS = Letter Number Sequencing Task (working memory); SSAT
= Simple Selective Attention; CSAT = Complex Selective Attention. Standardized coefficients are shown with significant coefficients
printed in bold. Z1 through Z3 are the residual variances of the dependent variables in the model. Reprinted with the publisher’s
permission.
Source: Willmott et al (2009). Reprinted with the publisher’s permission.

Spine Limb
–.12
.10
–.39
PTA 1
Employ-
.13 ment 1 yr
.11
–.34
.18
Mood 1
Educa- –.37
tion changes
–.15

.69 –.30

Pre-inj. .12 Cognitive 1

–.19 psych. .73
changes
disorder
.14
.26 .78

Pre-inj. Behavioral 1
employ. –.29 changes

–.24 .11 .08

Age Sex

FIGURE 4.3 Example of a full Structural Equation Model predicting outcome 1 year after traumatic brain injury.
Note: Final, modified model with improved model fit. The observed variables that indicate the mood, cognitive, and behavioral
change factors were entered into the Structural Equation Model but are omitted from this figure. Curved lines represent correlations;
straight lines represent regression coefficients. Correlations and standardized regression coefficients are displayed. PTA, posttrau-
matic amnesia.
Source: Originally published in Schönberger et al (2011b). Reprinted with the publisher’s permission.

Hammond_87543_PTR_04_18-22_08-12-14.indd 21 8/11/2014 11:00:19 AM

 22
■ In a first step, it was confirmed that the items
really were indicators of mood, cognitive, and
behavioral changes, respectively.
■ Once this was confirmed, the factor model
was entered into the full Structural Equation
Model displayed in Figure 4.3.
■ Whether this model was confirmed by the
observed data was then tested.
■ If a study is conducted at several sites (eg, a num-
ber of rehabilitation clinics) and associations between
variables are to be examined across all sites: Multilevel
Modeling (can be combined with Structural Equation
Modeling).
Applications
■ In general, as the name implies, studies of association
are concerned with relationships among variables.
■ In the context of clinical research, studies of asso-
ciation can be used for:
● Outcome prediction/responder analysis, for
example, as part of a clinical study
● Determination of the effectiveness of clinical
services under real-life conditions
■ For example, comparative effectiveness stud-
ies (often multicenter)
● Process research: (Temporal) association
between variables relevant to the clinical process,
such as the relationship between the quality of the
client–practitioner relationship and outcome
Hammond_87543_PTR_04_18-22_08-12-14.indd 22
I: DESIGN
When Not to Use
■ If the study goal is to determine the effectiveness or
efficacy of a new intervention, then randomized con-
trolled trials should be conducted, rather than studies
of association.
SUGGESTED READINGS
Hill AB. The environment and disease: Association or causa-
tion? Proc R Soc Med. 1965;58:295–300.
Kline RB. Principles and Practice of Structural Equation
Modeling. 3rd ed. New York, NY: Guildford Press;
2010.
Schönberger M, Ponsford J, Gould KR, Johnston L.
The temporal relationship between depression, anxi-
ety, and functional status after traumatic brain injury:
a cross-lagged analysis. J Int Neuropsychol Soc.
2011a;17:781–787.
Schönberger M, Ponsford J, Olver J, Ponsford M, Wirtz
M. Prediction of functional and employment outcome
1 year after traumatic brain injury: a structural equation
modelling approach. J Neurol Neurosurg Psychiatry.
2011b;82:936–941.
Shadish WR, Cook TD, Thomas D. Experimental and Quasi-
Experimental Designs for Generalized Causal Inference.
Boston, MA: Houghton Mifflin Harcourt; 2002.
Tabachnick BG, Fidell LS. Using Multivariate Statistics. 6th
ed. Boston, MA: Allyn and Bacon; 2012.
Willmott C, Ponsford J, Hocking C, Schönberger M. Factors
contributing to attentional impairment following trau-
matic brain injury. Neuropsychology. 2009;23:424–432.
8/11/2014 11:00:20 AM
 Brian Keogh, Jr, MD and Katherine W. Stenson, MD
Observational Studies:
Retrospective Versus
Prospective
DEFINITIONS AND DESCRIPTIONS
■ Retrospective: The population of interest is defined
after an outcome has occurred.
■ Prospective: The population of interest is defined
before an outcome has occurred.
■ Cohort: Group of patients, sharing a common char-
acteristic, assembled to analyze or observe over time.
■ Selection bias: When selection results in group dif-
ferences that may be related to and affect variations in
outcome. This results in a nonrepresentative sample
of the population.
■ Measurement bias: When knowledge of the presence
or absence of exposure to a risk factor or intervention
influences the assessment of disease or outcome on
follow-up.
■ Confounding: When a factor, other than the risk fac-
tor of interest, affects the outcome of a study. May
be accounted for statistically if the confounder is
recognized.
INTRODUCTION
■ The purpose of this chapter is to compare and con-
trast prospective and retrospective observational study
design with a discussion of the components, strengths,
and weaknesses, while providing examples of each
approach.
Hammond_87543_PTR_05_23-26_08-12-14.indd 23
5
BACKGROUND AND IMPLICATIONS
■ In both retrospective and prospective studies,
Subjects do not have the disease or outcome state pres-
ent prior to the baseline observation measurement.
■ Retrospective observational studies:
● A population is selected with a common starting
point, outcome, or timing of implementation of an
intervention.
● The required information to test the hypothesis
of interest is already available.
● Records are reviewed in order to gather the
data.
● Analyses are performed to determine what risk
factors may have contributed to the development of
an outcome or disease.
● Further data and follow-up are not obtained.
● For example, a study by Bennet et al (1999) ret-
rospectively examined the pooled chart data from a
cohort of 101 patients with complex regional pain
syndrome type I (CRPS I) who underwent spinal
cord stimulation (SCS) implantation between 1995
and 1998. The authors were able to conclude that
those who underwent implantation of a dual-lead
octapolar SCS with high frequency capabilities,
compared to those who had single-lead quadrapo-
lar systems, had a significantly greater reduction
in pain scores and surgical revisions to maintain
paresthesia coverage.
8/11/2014 11:00:35 AM
 24
■ This was a retrospective study because the data
were collected prior to defining the cohort.
■ Prospective observational studies:
● A sample is identified with a common expo-
sure and followed over time to assess for a given
outcome.
● Data are collected from the group during the
follow-up period.
● When the follow-up period is complete, the
group is analyzed to determine what risk factors
or exposures contributed to the development of the
given disease or other outcome.
● The analyzed data are obtained in the future.
● For example, in Wilson et al (2012), motor recov-
ery following traumatic spinal cord injury (SCI) in
84 patients was compared for those undergoing
surgery less than 24 hours after injury and those
who underwent surgery more than 24 hours after
injury. There was a statistically significant positive
effect associated with early surgery with regard to
American Spinal Injury Association (ASIA) motor
score after adjustment for injury level and preop-
erative state.
■ This is an example of a prospective study
because the cohort was defined and recruited
into the study prior to data collection.
■ Practice-based evidence (PBE) study design:
● This is a prospective observational study design
that was developed to assess populations that
would be more representative of the heterogene-
ity encountered in real medical practices. While
traditional prospective observational studies con-
trol for variability through randomization and by
excluding patients from the cohort, the PBE design
allows for unlimited variability and potential con-
founders. The heterogeneity in the sample and
interventions are accounted for and accurate mea-
surement of all variables is obtained. The influence
of these variables can then be determined by using
a multivariate statistical analysis with weighting of
the clinical importance of each factor.
● The PBE design also allows for the analysis of
the contributions of multiple interventions at the
same time, such as is typically encountered in stan-
dard practice when a patient is being treated with
multiple interventions simultaneously (see articles
by Horn and Gassaway, 2010).
STRENGTHS
■ Retrospective:
● No loss to follow-up, since the cohort is assem-
bled from available data.
Hammond_87543_PTR_05_23-26_08-12-14.indd 24
I: DESIGN
●Time efficient, since the measurements have
already occurred.
● Cost effective.
● Able to establish predictor variables of the
outcome.
● Reduced chance of bias during measurements as
the research question and expected outcome were
not known.
● Good method for studying rare diseases.
■ Prospective:
● Allows for more accurate data collection since
the determination of which variables to measure is
made prior to the outcome, allowing customization
of follow-up.
● Considered higher quality evidence as compared
to retrospective observational studies because study
conditions can be controlled prospectively.
● A strong inference of association can be made
due to the time sequencing of events, that is, the
measurement occurs prior to the outcome.
● Valuable for studying fatal outcomes, since measure-
ment of variables are obtained directly from the research
participant rather than from records or a proxy.
■ For example, the quantity of cigarettes smoked
by the participant is monitored as opposed to
family members’ retrospective report of the
subject’s cigarette use.
■ Both retrospective and prospective observational
studies:
● Allow study of numerous variables and out-
comes simultaneously.
● Are often more generalizable to clinical practice
than randomized controlled trials, due to broader
inclusion and fewer exclusion criteria, that is,
potentially more representative samples.
WEAKNESSES
■ Retrospective:
● Only use available data from the past without
control over the nature or quality of the measure-
ments; thus, important data may not have been
recorded.
■ Prospective:
● Is time inefficient since long follow-up is typi-
cally required.
● There is greater cost associated with following for
a long time, measuring large numbers of variables,
and often requiring a large number of subjects.
● It is difficult to study rare diseases due to lack
of numbers.
● Prone to high loss of follow-up.
● Prone to selection bias and measurement bias.
8/11/2014 11:00:35 AM
 5: OBSERVATIONAL STUDIES: RETROSPECTIVE VERSUS PROSPECTIVE
●May miss preclinical evidence of a disease that
is already there.
■ For example, a stroke patient with nonclini-
cally evident coronary artery disease (CAD)
that limits participation in exercise. In a study
investigating the effects of exercise on develop-
ment of CAD, it may appear that lack of exer-
cise is a risk factor for CAD, when it may be
that the lack of exercise is a result of CAD.
■ Both retrospective and prospective observational
studies:
● Are susceptible to the effects of confounding,
making causal effects difficult to establish.
■ To some degree, this may be addressed by pro-
pensity scoring or sensitivity analysis (see
Chapter 6).
STRATEGIES
Sample (See Table 5.1)
■ Retrospective:
● Observational cohort with the exposure point
and outcome in the past, that is, follow-up time is
past-to-present.
● The data of interest were obtained in the past
and are analyzed in the present. There is no further
follow-up on the subjects.
■ Prospective:
● Observational cohort with the exposure point in cur-
rent time, that is, follow-up time is present-to-future.
● The data are accumulated during a set follow-up
period.
● Useful to select cohorts with a high incidence of
the outcome of interest to minimize the required
numbers.
■ Both retrospective and prospective:
● The time sequence between exposure to risk fac-
tor and development of outcome is established.
TABLE 5.1 Comparison of Retrospective and Prospective Observational Studies With Regard to Timing of Data
Collection, Cohort Creation, and Data Analysis
Past
Retrospective Common outcome, exposure or Cohort assembled
intervention Data analyzed
Data gathered
Prospective Common outcome, exposure or
intervention
Cohort assembled
Initial data gathered
Hammond_87543_PTR_05_23-26_08-12-14.indd 25
25
Controls
■ Retrospective:
● The selected control is a random group of sub-
jects without the exposure/risk being studied who
are matched to be demographically similar to the
sample of interest.
■ Prospective:
● Internal control: A group that is selected from
the followed subjects that either did not have the
exposure/risk factor or had a different level of risk/
exposure.
■ The control is not defined until after the
cohort has been assembled and measurements
have begun.
■ This type of control is often used in population
study cohorts, where the subjects from a given
locale are followed, such as in the Framingham
study.
■ For example, a prospective cohort of
patients who have had total hip arthroplasty
(THA) is assembled to study the likelihood of
mortality following THA when a myocardial
infarction occurs within 4 weeks after sur-
gery. After 4 weeks there would be patients
in the cohort who had a myocardial infarction
and of those, there would be a subgroup who
experienced the outcome of interest (mortal-
ity) and a subgroup who did not. There would
also be a subgroup of patients who did not
have a myocardial infarction in the 4-week
study period. The subgroup without myo-
cardial infarction would serve as the internal
control.
● External control: A group that is demographi-
cally similar but without the exposure of interest. If
opioid use and risk of motor vehicle crash were to
be studied, an appropriate external control would
be a group that does not use opioids but is other-
wise similar.
Time
Present Future
Further follow-up obtained
Data analyzed
8/11/2014 11:00:35 AM
 26
■ When the group to be studied is too homoge-
neous with regard to exposure/risk factors, an
external control can be selected.
■ For example, a cohort of patients with acute
stroke who present to an inpatient rehabilitation
facility is assembled to determine the likelihood
of discharge home versus discharge to a nursing
facility, based upon the presence of urinary incon-
tinence. An external control would be a group of
similar stroke patients admitted to acute inpatient
rehabilitation without urinary incontinence.
Applications
■ Retrospective:
● For questions that can be addressed using avail-
able data.
■ Prospective:
● Useful for diseases or outcomes that are rela-
tively frequent.
● When less expensive methods have failed to
answer the question of interest.
When Not to Use
■ Retrospective:
● Required data to answer the question are not
available.
Hammond_87543_PTR_05_23-26_08-12-14.indd 26
I: DESIGN
■ Prospective:
● Rare diseases or outcomes.
● Time and resources are limiting.
SUGGESTED READINGS
Bennett DS, Aló KM, Oakley J, Feler CA. Spinal cord stim-
ulation for complex regional pain syndrome I [RSD]: a
retrospective multicenter experience from 1995 to 1998 of
101 patients. Neuromodulation. 1999;2(3):202–210.
dos Santos Silva I. Cancer Epidemiology: Principles and
Methods. France: International Agency for Research on
Cancer; 1999.
Euser AM, Zoccali C, Jager KJ, Dekker FW. Cohort stud-
ies: prospective versus retrospective. Nephron Clin Pract.
2009;113(3):c214–c217.
Horn SD, Gassaway J. Practice based evidence: incorpo-
rating clinical heterogeneity and patient-reported out-
comes for comparative effectiveness research. Med Care.
2010;48(6 Suppl):S17–S22.
Horn SD, Gassaway J. Practice-based evidence study
design for comparative effectiveness research. Med Care.
2007;45(10 Suppl 2):S50–S57.
Hulley SB, Cummings SR, Browner WS. Designing Clinical
Research: An Epidemiologic Approach. Baltimore:
Williams & Wilkins; 1988.
Wilson JR, Singh A, Craven C, et al. Early versus late sur-
gery for traumatic spinal cord injury: the results of a pro-
spective Canadian cohort study. Spinal Cord. 2012;50:
840–843.
8/11/2014 11:00:35 AM
 Whitney Pratt, MD, PhD and Katherine W. Stenson, MD
Historical Controls
DEFINITIONS AND DESCRIPTIONS
■ Historical control group:
● An external control group made up of persons
who have been previously studied and who are
similar in characteristics to the present sample.
● Nonconcurrent and nonrandomized control.
INTRODUCTION
■ This chapter reviews the appropriate use of histori-
cal controls in the research setting, identifies the limi-
tations of such controls, and discusses strategies for
using them successfully.
IMPLICATIONS
■ Historical control groups provide unique opportuni-
ties to examine research questions when more stan-
dard control groups are unavailable.
■ Understanding the proper application of this type of
control will permit more effective study design and
data interpretation.
BACKGROUND
■ Historical controls have traditionally been used
in medical research in situations where securing an
acceptable contemporary control group is impossible
or impractical.
Hammond_87543_PTR_06_27-29_08-12-14.indd 27
6
■ However, historical controls are often used inap-
propriately. Researchers should give careful thought
to the inclusion of this type of control group to ensure
the suitability of this method for their project.
STRENGTHS
■ Rapid
■ Inexpensive
■ Avoids need to withhold potentially beneficial treat-
ment, which may be unethical
■ Improved recruitment due to all participants receiv-
ing potentially beneficial therapy
WEAKNESSES
■ Tend to exaggerate the value of a new treatment
■ Vulnerable to bias
■ Large differences on covariates (eg, age and gen-
der) could lead to biased estimates of treatment effect
between the treatment and control groups
■ Changes in outcome over time may be due to
changes in various factors other than the introduction
of the intervention being studied
● Underlying patient populations
■ Recruitment of patients, for example, from
dissimilar patient care settings or cultures may
introduce confounding variables.
● Criteria for selecting patients
■ Different inclusion/exclusion criteria may
create disparate control and treatment groups,
8/11/2014 11:00:49 AM
 28
for example, in disease severity, comorbidities,
or concurrent treatments.
● Patient care and management peripheral to
treatment
■ Changes in treatment of the primary condi-
tion or comorbidities may incidentally affect
outcomes.
● Diagnostic or evaluating criteria
■ Using different assessment tools to measure
independent/dependent variables or outcomes
may lead to an inability to accurately compare
groups.
● Quality of data available
■ Variations in how or when data are recorded
may impact the ability to accurately compare
groups.
● Example: A proposed study of the effect of spi-
nal cord injury rehabilitation on functional out-
comes of paraplegic veterans
■ Historical controls: 30 paraplegic veterans
who had recently completed a different study
conducted at the same facility who only received
physical therapy.
■ Weakness: No other obvious systematic
changes in patient care between times in which
patient groups were evaluated, but there may
have been unrecognized differences.
STRATEGIES
Requirements for a Valid Historical
Control Group
■ A relatively recent study or data are available for
same indication
■ Subject samples with similar demographics and
prognostic factors
■ Similar inclusion/exclusion criteria
■ Similar study procedures or standard of care
■ Similar outcome measures
Sources of Historical Control Data
■ Literature:
● Subject to publication bias
● Usually not accepted by the FDA
■ Database:
● Patient-level data from previous study, medical
records, or other database
● More powerful than literature reference, but not
always available
Hammond_87543_PTR_06_27-29_08-12-14.indd 28
I: DESIGN
Statistical Adjustments Used to Limit Bias
or Overestimation of Effects
■ Matching controls to subjects:
● Example: Proposed study on the effect of
robotic-assisted gait training on the neurological/
functional outcomes of patients with subacute spi-
nal cord injury
■ Historical controls: Patients treated in the
same department in previous years, matched to
current patients based on age, severity of injury,
level of injury, and cause
■ Covariate adjustment:
● Limited in the number of variables that can be
included, although propensity analysis provides
method to include a large number of covariates in
studies with a large sample (see below)
■ Interpretation can be challenging
■ Bayesian methods
■ Propensity scores (see D’Agostino, 1998):
● Provide an estimate of the effect of an interven-
tion by accounting for the covariates that predict
receiving the treatment, thereby reducing selection
bias by equating groups based on these covariates
■ Sensitivity analysis (see Greenland, 1996):
● Used to determine what impact any unmeasured
covariates would have to have on the data in order
to alter the conclusions of a study
When to Use
■ Risk of withholding treatment is too high:
● If an effective treatment exists, use of a placebo
control may not be ethical
● Example: Proposed study of prevention of aspi-
ration pneumonia in stroke patients at an inpatient
rehabilitation facility using a newly implemented
clinical pathway for the evaluation and treatment
of dysphagia
■ Historical controls: Patients treated at same
facility the year prior to implementation of the
dysphagia protocol.
■ Results: No patients on the clinical pathway
developed aspiration pneumonia during the period
of the prospective study versus 6.7% who devel-
oped pneumonia during the pre-pathway year.
■ No other treatment exists that is appropriate to use
as a control.
■ Rare conditions for which recruitment may be
difficult.
■ The untreated condition is well understood and
characterized.
8/11/2014 11:00:49 AM
 6: HISTORICAL CONTROLS
■ Outcomes of a standard treatment (or no treatment)
are well known and vary little for a given population.
■ The experimental therapy is expected to have a very
large beneficial effect.
● Use of a historical control group avoids with-
holding of potentially effective treatment.
● Limits concern regarding exposure of all sub-
jects to experimental treatment while allowing the
focus of study to remain on primary outcome.
When Not to Use
■ No undue risk associated with receiving placebo
versus experimental therapy
● Unless withholding an experimental treatment is
felt to be unethical due to risks associated with not
receiving the treatment, a more rigorous placebo-
controlled study is preferable.
Hammond_87543_PTR_06_27-29_08-12-14.indd 29
29
■ Alternative therapies could serve as a suitable pro-
spective control treatment.
■ No appropriate historical sample exists.
■ Use of placebo will not negatively impact subjects.
SUGGESTED READINGS
Baker SG, Lindeman KS. Rethinking historical controls.
Biostatistics. 2001;2(4):383–396.
D’Agostino RB Jr. Propensity score methods for bias reduc-
tion in the comparison of a treatment to a non-randomized
control group. Stat Med. 1998;17(19):2265–2281.
Friedman LM, Furberg CD, DeMets DL. Fundamentals of
Clinical Trials. 4th ed. New York: Springer; 2010.
Greenland S. Basic methods for sensitivity analysis of biases.
Int J Epidemiol. 1996;25(6):1107–1116.
Yoshimura I, Matsumoto K. Notes on the use of histori-
cal controls. Environ Health Perspect. 1994;102(Suppl
1):19–23.
8/11/2014 11:00:49 AM
Another random document with
no related content on Scribd:
significantly told me, of the King’s own pasturage. There was nothing
to be done but to accompany them; so telling some of the Hy
Soumaulee to come to Farree the next morning to see me, and if I
were not there to go on to Ankobar, I proceeded with my guides, or
guards, to the same house I slept in the last night; and the ready
smiling welcome, the little bustle to receive me cordially, I met with
from the good-natured inmates, was some set-off to the brutal
indifference of the state-gaoler; for such office also I found was filled
by the head of the customs of Shoa, the Abigass, or frontier governor
of Efat, the obsequious spiteful pluralist the Wallasmah Mahomed.
I passed the night, having received no answer to my note from
Ankobar, wishing for the day, still hoping that I might be mistaken in
my fears, and that some of the members of the embassy would
come to congratulate me on my safe arrival, and free me from the
anxiety, restraint, and espionage I was now annoyed with; for two
sentinels were constantly on duty in the little enclosure, and always
present in the house, when I received visits even from my Hy
Soumaulee friends.
The next day came, but no news from Ankobar. I amused myself
as well as I could, writing up my notes and settling small accounts
with my escort and those of the Kafilah people, from whose
importunities on the road I had relieved myself by promises of
presents in Shoa, and who now came for paper, needles, buttons,
scissors, and razors. Almost all that I possessed was divided among
them as some little return for their continued kindness and fidelity to
me on the road; for I had little to complain of except the continual
falsehoods and petty deceits practised invariably by the Tajourah
people. Even Ohmed Medina was not altogether exempt from this
failing; but it was from a motive of well-meant kindness, so that I
should not be able to detect the number of instances that little
attempts were made to impose upon me, and which he thought
might lead to expostulation and angry discussions.
 CHAPTER II.
Detained at Farree.—​No news from Ankobar.—​Fear all is not right.—​
Escape from my confinement.—​Reach Garcia Mulloo.—​Followed
by officers of Wallasmah.—​Compromise matters.—​Return to
Farree.—​Brutality of Wallasmah.—​Planning escape to the coast
with Hy Soumaulee.—​Arrival of Mr. Scott from Ankobar.—​Chief
cause of my detention.
I stayed in Farree anxiously awaiting some news from the embassy,
until the 25th; but neither note nor messenger came to relieve the
suspense I was in. The night before, Ohmed Medina, however, had
called upon me, and told me that all was right as regarded their
personal safety, but informed me that my note from the Dinnomalee
had been intercepted by the Wallasmah, and that none of the
English in Shoa knew that I was in the country. I made up my mind,
on hearing this, to attempt getting to Ankobar the next morning, if it
were possible; and accordingly, before it was light, opened the little
wicket that served for the door, passed unobserved the two sentinels
who lay wrapped up in their body cloths fast asleep, and was soon
some distance on the wrong road; that is to say, the most circuitous
one to Ankobar. I thought that I was not exactly right, and meeting
some labourers going into the fields to work, I asked the way, by
repeating the word, Ankobar. They were too much surprised to
speak, but pointed in the direction of the road, and I left them staring
after me with a wondering look, as if to ask what would come next.
Having reached a village about five miles to the north-west of Farree,
I found it impossible to go on, it having been one continual ascent
along the roughest and most winding path that can well be imagined.
Oppressed with difficulty of breathing, fatigued, and foot-sore, I
turned toward the door of the first house, and sitting down on a
stone, made signs that I wanted some water. Hereupon such a
screaming was set up by the only inmates, two naked children, that it
could not have been exceeded if I had intimated that they were
about to be devoured. Their cries brought two other little girls, who
came running round the house, but seeing me, promptly turned
back, tumbling over each other to get out of the way, contributing as
they lay not a little to the frantic roaring of the children inside.
The noise soon brought all the disposable people, men and
women of all ages, who had not left the village for their labours in the
fields, who soon recognised in their visitor a Gypt or Egyptian, as the
Abyssinians call all white men. I was glad to find that the character
seemed to be a very respected one, although the first evidence I had
of it, was the numerous beggars for articles, the names of which I did
not understand. They invited me into the house out of the sun, and a
large wooden mortar was laid on its side for me to sit upon, whilst
several women employed themselves scorching some coffee beans,
in a coarse earthenware saucer over a little wood fire in the centre of
a circular hearth, that occupied the middle of the room. The whole
house consisted of this one apartment, the surrounding wall being
composed of sticks placed close together, and about four feet high,
upon which rested a straw thatched roof frame of light bamboo, well
blackened with the smoke.
I had not long arrived at Garcia Mulloo, the name of the village,
before I was followed by a large body of men armed with spears and
staves, and dragging along with them, most unwillingly, my old grey
mule. The misselannee of Farree, whom I knew, was at the head of
the party, and appeared very well pleased to see me, addressing me
with great politeness, though I could not understand a word that he
said. I took care, however, in Arabic, to charge him and the
Wallasmah with incivility, and want of hospitality, for detaining me so
long in Farree against my will, and also with having, like a thief,
stolen the note I had sent to Ankobar. As we had been now joined by
a man named Brekka, who understood what I said, he interpreted for
us, and afforded the misselannee an opportunity in reply, of throwing
the whole blame upon the Wallasmah, whose servant he was, at the
same time begging me to return with him, for which purpose, and for
my accommodation, he had brought my mule along with him. I
positively refused, on the plea, that their King had promised mine,
that Englishmen were to travel unmolested through the country,
alluding to the treaty, and that, accordingly, if they now used force to
take me back to Farree, it would bring the matter to an issue, and my
people would then see the real value of the word of Sahale
Selassee. Seeing I was determined not to return with them they
agreed to compromise the matter, upon my promising to remain at
Garcia Mulloo, and not attempt to proceed farther towards Ankobar,
until the King’s pleasure respecting me should be known. This I was
induced to do by the misselannee’s pacific appeal that I would not do
anything which would occasion him to be imprisoned, and all his
property confiscated.
Our interpreter, Brekka, was a scamp of a renegado, who had
been a Christian, but was converted to the Islam faith, by the
promise of a situation under the Wallasmah, whose district, the
province of Efat, is inhabited chiefly by Mahomedans. The contiguity
of the two faiths among a people of one origin, affords an interesting
opportunity of examining the first effect of differences in religious
belief, and which leads, in the course of time, to the division of one
family of man into two distinct nations, differing in customs, pursuits,
and even, after a lapse of time, in physical features.
The same dispersing operation of opinion, but more advanced, is
to be observed in the separation, at the present day, of the Dankalli
and Soumaulee tribes, and to any zealous student of the science of
all sciences, humanity, or the natural history of man, it is
indispensably necessary that he should visit the countries of
Abyssinia, of Sennaar, and Adal, where he will find collected, as at a
centre, the originals of all the different varieties into which
physiologists have divided the human race; and where, at this
moment, the principal causes of the great moral change in the
condition of man, consequent upon the flood, may be observed in full
operation, and producing the same effects of dispersion. Christian
civilization, which points to a future union, is the antagonizing
principle to the opinion disturbing one, which, I believe, alone
separates and divides mankind; and I could wish to see, here, in
intertropical Africa, a Mission of enlightened ministers of the Gospel,
whose object should be simply to spread the easily understood
doctrine of one God, and that love and truth are the redeeming
principles in the character of man, to restore him to that state of
excellence from which he has fallen.
It being arranged I should stay at Garcia Mulloo, a supply of bread
and beer was ordered by the misselanee, who had been sent for to
see after this duty; the same officer of the town of Farree, returning
there with his party, taking my mule along with them, and leaving
Brekka and another man to keep me company, as was said, but in
fact, to keep guard over me. A disjointed conversation with the
former served to amuse me during the rest of the day. He gave me
some information respecting the Embassy, and of the dislike
entertained by Sahale Selassee to the English; which surprised me
considerably, nor would I at first believe it, but ascribed the
statement to the ill feeling and jealousy with which the visit of our
Political Mission to the Court of Shoa, was viewed by the
Mahomedans of Efat.
In the afternoon, Brekka walked down to Farree, and when he
returned, told me he had seen a letter for me, and a messenger from
Ankobar, and that if I wished to see them I must go to that town. I did
not hesitate a moment, but was now as anxious to be off, as I was
before obstinately bent upon remaining. The news of Brekka being
confirmed by the arrival of a messenger from the Wallasmah, with
the same information, I started immediately. I conceived that the not
sending the letter to Garcia Mulloo, was perhaps intended as a kind
of punishment for my breaking prison in the morning.
In about an hour and a-half, we were again crossing the little
stream which flows at the base of the hill on which Farree stands;
and I was soon seated in my old quarters, whilst Brekka went to
obtain for me the expected note. When he returned, he brought me
an order to go to the Wallasmah myself, as he wanted to see me;
and who occupied a house upon one of the little eminences opposite
to mine. I was not long in presenting myself in obedience to this
summons, and found that gentleman sitting upon a large oxskin
spread upon the ground, paring his toe nails with an old pocket knife.
As I came round the low stone fence against which he leaned, he
cast his eyes upon me, and growled a very sinister kind of salutation,
asking me in broken Arabic how I did. I now requested him to give
me the letter from Ankobar, but he only shook his head. I asked to
see the messenger, and with a chuckle of triumphant cunning, he
pointed with the open knife to the fastened door of an outhouse, an
action which I readily interpreted to mean, “He is there, in prison.” I
did not say a word more, but walked away in high dudgeon,
overturning a rude Abyssinian who, with spear and shield pushed
against me, as if to prevent my exit when I made my way out through
a little wicket in the stick enclosure that surrounded the house.
The worst of my situation was, that I had no friends near, all the
Hy Soumaulee and Tajourah people being according to custom,
obliged to locate themselves during their stay in Shoa, in a little town
called Channo, situated about two miles to the north-east of Farree,
where they are compelled to leave their shields and spears when
they go farther into the interior of the country. I had to send for any of
these to come to me, but either it was too late in the day, being after
sunset, or orders had been issued to the contrary, for I could induce
no one to take a message from me either to Ohmed Medina or
Carmel Ibrahim. I was obliged, therefore, to remain quiet for the
night, being determined, however, on the morrow to escape into the
Adal country, and carry the news back, which otherwise might be a
long time in reaching Aden, of the actual condition of things in Shoa;
where, instead of the English being courted and caressed, as was
believed to be the case when I was in Tajourah, they were, in fact,
the objects of the most jealous suspicion, and subjected to the most
tyrannical surveillance, if not actually in prison.
The early part of the morning of the 26th of May was occupied in
projecting the plan of my escape with Carmel Ibrahim and Adam
Burrah, the latter of whom having assisted Lieut. Barker in getting
through the Adal country after that gentleman had left the Hurrah
Kafilah, I could the more confidently rely upon, although I had not the
least doubt of the fidelity or trustworthiness of the former. These two
had come with Allee the First to laugh with me at my attempt of
getting to Ankobar the day before, and endeavoured to soothe and
interest me, as they thought, by showing how they would
disembowel the old fat Wallasmah if they had him in their country.
My proposal to go back was met with their decided approbation. It
was accordingly arranged that Carmel and Adam should accompany
me in the evening, whilst the rest of the escort were to remain, and
during the night manage to steal my mule, and as many more as
they could, and join us at the little Wahama town Dophan, beyond
which they knew very well no attempt would be made to pursue us.
I was in the act of making a few cartridges for my anticipated
return journey, when I heard a loud cry of “Commander,
Commander,” an English word, by which the Abyssinians had been
taught to designate the head of the Mission. Two or three of the
inhabitants of Farree came also in a great hurry to call me out of the
house, and tell me that some Gypt or other was approaching. I was
equally eager, and even ran in a most undignified manner to meet
this messenger of light, who, mounted on a mule, now appeared
upon the summit or crest of the road before it descends into the little
hollow where stands the market-place. There was a great air of
civilization about him. He wore a broad-brimmed hat, somewhat like
my own, covered with white cotton cloth, a sailor’s large pea-jacket
belted round his waist, an old pair of grey check trowsers, and came
with a sober steady pace along the narrow path.
I met him as he dismounted beneath the few mimosa-trees, and
after a hearty shake of the hand, invited him to my hotel. He then
introduced himself as Mr. R. Scott, the surveying draughtsman
attached to the Mission.
His first explanation was the cause of his non-arrival sooner,
which was owing to the utter ignorance of my arrival on the part of
Captain Harris, the chief of the Embassy, until the night but one
before, when the King had forwarded by one of his pages two notes,
which I had endeavoured to send to him, the last one dated from
Dinnomalee. The other was the one which had been sent by Esau
Ibrahim, who, it will be remembered, was despatched from Mullu, on
the other side of the Hawash, with a note to Ankobar, informing
Captain Harris of my being on the road with stores. Both these
letters had been intercepted and detained, until public rumour
spreading, the King could not have kept the Embassy much longer
ignorant of my being in the country; and he therefore made a virtue
of necessity, and sent the letters before they were demanded.
 An answer had been sent to me by Capt. Harris the day before by
the messenger now in prison, confined by the Wallasmah for having
brought a letter for me, after the King had issued orders that all
correspondence between the English already in the country and
those arriving should be prevented. Mr. Scott was not at all surprised
when I informed him of the circumstance, though I certainly
considered such a proceeding to be very much at variance with the
conditions and stipulations I was given to understand were contained
in the commercial treaty. I could not help remarking this, and Mr.
Scott then candidly admitted the King did not know the character or
purport of the paper he had signed; and had only been made aware
of the new responsibilities he had incurred, by a sharply worded
expostulatory letter, written by Mr. Krapf, in accordance to the
dictation of Captain Harris, on an occasion subsequently to the
signing of the treaty, when despatches and letters coming up from
the coast were intercepted and detained for some time by the orders
of the King. Singularly enough, this information was corroborated by
Ohmed Medina, who told me that my letter from Dinnomalee had not
been carried to Captain Harris, but to the King, who wanted to find
out whether the English were his friends or not, and was trying my
disposition and that of the Commander (Captain Harris) by this harsh
treatment of me; a kind of experiment, in fact, to see what would be
borne by us, and how far he had limited his authority by attaching his
signature to the treaty. Any idea of granting public benefit, at the
expense of his prerogative was never entertained for a moment, the
intentions of the King being limited to shewing personal favour alone,
which he is ever ready to concede even now to English travellers,
much as he complains of the conduct of the Mission in Shoa as
regards their political misdoings; more especially of the great insult
offered to him by the unfortunate letter before alluded to, and which
was worded so unguardedly, that the King, on receiving it, might
well, considering his great regard for Mr. Krapf previously, turn to him
and say, in a tone that implied more of sorrow than of anger, “Did
you write that, my father?”
 CHAPTER III.
Staying at Farree with Mr. Scott.—​Both placed under parole.—​
Description of the houses of Farree.—​Of the flour mill.—​Some
remarks upon the origin of the Amhara.—​Dr. Prichard upon
identity of the Amhara with the Antomali of Herodotus.—​Physical
characters of the people.—​Interview with the Wallasmah.—​
Saltpetre rock.—​Province of Efat.—​Take leave of Escort.—​
Tyrannical conduct of the Wallasmah.
May 26, 1842.—After Mr. Scott joined me at Farree, I considered
that all my troubles were at an end, although I had still to go above
fifty miles before I could meet the members of the British Political
Mission who had accompanied the King to his residence at
Angolahlah, the most western town of his dominions. An
establishment was still kept up at Ankobar, situated about one third
of the way between Farree and Angolahlah, at the head of which
was the naturalist attached to the Mission, Dr. Roth, with whom was
Mr. Bernatz, the artist, and there also Mr. Scott was stationed.
Captain Harris the Ambassador, Captain Graham, the second in
command, and Mr. Assistant-Surgeon Kirk, lived at Angolahlah,
where I now expected to be permitted to go by my gaoler the
Wallasmah. I found, however, I was reckoning without my host, for a
new difficulty had arisen, from the circumstance of Mr. Scott having
come down to Farree without the permission of Walda-anna, the
Governor of Ankobar. He was accordingly given to understand that
he must consider himself a prisoner with me until the pleasure of the
negoos should be known as to our disposal. It was in vain we
expostulated with our surly gaoler; we were to be opposed by force if
we attempted to leave Farree, and other sentinels were charged with
the care of us. Something we did effect, and that was the liberation
of the messenger who was detected bringing me a letter the day
before, for as soon as this request was made to the Wallasmah it
was at once acceded to, and the man was ordered to be set at
liberty. Taking this as an evidence of some relaxation of the harsh
treatment with which we had been treated, we sat sometime chatting
with the old gentleman, and I hinted my intention of making him
some present if he would honour me so far as to accept of my poor
gifts. When we got up from the ground where we had been sitting,
the Wallasmah directed his son, a fine young man about three or
four and twenty years old, to accompany us to our residence; a
sufficient intimation of his being graciously disposed, without the
broad hint given by one of his followers, who whispered into the ears
of Mr. Scott, “Give your memolagee to that man.” Our imprisoned
servant not making his appearance before we left the Wallasmah, we
asked where he was, and were surprised to hear that he had left
Farree for Angolahlah without seeing us, but which we supposed he
had been obliged to do, so that there should be no chance of our
slipping a note into his hands for our friends in that town.
We returned to our house, and for the rest of the day amused
ourselves with hearing and telling whatever most interested us,
whether of home or foreign news. I must observe that a present of
three pieces of calico and a pound of gunpowder was made to the
Wallasmah, who sent us back his compliments, and that he was
highly delighted with the present, but would be obliged for a little
more gunpowder.
Mr. Scott and I were entertained and taken care of for four days in
Farree, much to our discomfort and vexation. Fortunately this
gentleman had brought with him two native servants, who made
themselves useful by marketing and cooking during the term of our
confinement, so we suffered nothing from want of food. We could
also walk about the straggling town on pledging our word that we
would not attempt to escape, although our parole was not deemed
sufficient, for, like Buonaparte at St. Helena, two sentinels, on such
excursions, always followed at a certain distance in our rear.
Many of the houses in Farree, instead of being the usual circle of
closely placed sticks, some five feet high and surmounted by a high
conical straw roof, are partial excavations in the soft trachytic stone,
so as to leave a back and sides of natural rock. Over this is laid a flat
roof, consisting of untrimmed rafters covered by a thick layer of
brushwood, upon which is placed a layer of earth some inches in
thickness, well stamped down with the feet. A front of wattled sticks,
in which the entrance is made, completes the house, and in one
such as this was I lodged during my stay in this town.
The internal arrangements were equally simple. A raised platform
of stones and clay, about two feet high, occupied one half of the
single apartment, and upon one end of this, reaching to the roof,
stood a huge butt-like basket, smoothly plastered over inside and out
with clay. This was the family granary, in which was preserved the
teff seed, or wheat, from the depredations of the numerous mice that
are a thorough pest in Abyssinia. In a corner below, stood side by
side two of the peculiar handmills used in this country, each
consisting of a large flat stone of cellular lava, two feet long and one
foot broad, raised upon a rude pedestal of stones and mud, about
one foot and a half from the ground. The rough surface of this stone
sloped gradually down from behind forwards into a basin-like cavity,
into which the flour falls as it is ground. A second stone, grasped in
the hand of the woman who grinds, weighs about three pounds,
beneath which, as it is moved up and down the inclined plane of the
under millstone, the grain is crushed, and gradually converted into a
coarse flour.
This is the same kind of mill that was used by the ancient
Egyptians, and is represented in the excellent work upon those
people, recently written by Sir G. Wilkinson, although he describes it
as being used for fulling clothes, having mistaken, I suppose, the
flour represented as falling into the cup-like recipient for a stream of
water. I observe, also, in another plate in the same work, a
representation of this mill, but without any allusion to its real
purposes. Moses, in the fifth verse of the eleventh chapter of
Exodus, describes exactly the character of the occupation, and the
instrument, where he speaks “of the maid-servant that is behind the
mill,” for women are only employed on this duty, and they always
stand in the rear, leaning forward over their work. Very few houses,
those only of the poorest people, have but one mill; generally two or
more stand side by side in a row, and the number is always
mentioned when the idea is wished to be conveyed of the large
dependent retinue that the master of the house feeds.
A few large jars containing water, or ale, ranged along one side of
the house, and a shield hung from the projecting end of one of the
sticks that formed the front, were the only articles that occupied
prominent positions as furniture in my residence. Three or four
“maceroitsh,” or earthenware pots for cooking, generally lay upset in
the white wood ashes contained in the large circular hearth that
occupied a portion of the floor opposite to the mills; and some of the
necessary but small instruments for clearing or spinning cotton were
placed when not being used upon a skin bag, in which a quantity of
that useful material was contained.
I was very much struck with the extreme contrasts that could be
drawn between the inhabitants of Farree and the Dankalli Bedouins.
The large and portly forms of the former, their apparent love of quiet,
the affection they evinced for their children, and that of the children
for their parents, were all points characteristic of these great
differences. The physiognomy of the two people exhibited equally
varying features, and as the men of Farree are a good type of the
real Amhara population, I shall endeavour to give an idea of the form
of the countenance and the head peculiar to this family of man, by a
description drawn from my first observations in that town, where the
people have less admixture of Galla blood, than the inhabitants of
the table land of Shoa above and beyond them.
This will be preceded, however, by some necessary, and, I
believe, novel information respecting the origin of the Amhara, which
I became acquainted with during my residence in Shoa, and which
has been singularly confirmed by a comparison of the reports and
prejudices I noted down while in that country, with recorded
circumstances of the earlier history of Egypt, and of other powerful
empires that once existed along the course of the Nile.
Amhara, which word is at present only used to designate the
Christian population of Abyssinia, was, previous to the introduction of
the Mahomedan religion, the descriptive appellation of an extensive
red people, who principally occupied the eastern border of the
Abyssinian table land, from the latitude of Massoah in the north to
that of lake Zui in the south. To the west of these, and occupying the
portion of the table land in that direction, lived a people decidedly
different in their complexion, their features, their language, their
religion, and their customs. These were the Gongas, or Agows, who
I believe to have been the original possessors of the whole plateau,
until a period remarkable in history, when the Emperor of Meroë or
Ethiopia located upon a portion of their country, those disaffected
soldiers of Psammeticus who had sought an asylum in his kingdom.
Were I not convinced that the Amhara population of Abyssinia, at the
present day, can be physically demonstrated to be the descendants
of these fugitives from Egypt, I would not venture to advance such
an innovation upon the generally received opinion, that the Amhara
are aborigines of the country they now inhabit.
Under the term Abyssins, Dr. Prichard, in his invaluable work
upon the natural history of man, includes all the different nations that
now inhabit the lofty plain of Abisha or Abyssinia. Of one of these
nations, the Amhara, he remarks, “So striking is the resemblance
between the modern Abyssinians and the Hebrews of old, that we
can hardly look upon them but as branches of one nation, and if we
had not convincing evidence to the contrary, and knew not for certain
that the Abramidæ originated in Chaldæa, and to the northward and
eastward of Palestine, we might frame a very probable hypothesis,
which would bring them down as a band of wandering shepherds
from the mountains of Habesh, and identify them with the pastor
kings, who, according to Manetho, multiplied their bands in the land
of the Pharaohs, and being, after some centuries, expelled thence by
the will of the gods, sought refuge in Judea, and built the walls of
Jerusalem. Such an hypothesis would explain the existence of an
almost Israelitish people, and the preservation of a language so
nearly approaching to the Hebrew in intertropical Africa.” The
learned ethnologist goes on to observe—“It is certainly untrue; and
we find no other easy explanation of the facts which the history of
Abyssinia presents, and particularly of the early extension of the
Jewish religion and customs through that country, for the legend
which makes the royal house of Menilek descend from Solomon and
the Queen of Sheba, is as idle a story as ever monks invented to
abuse the reverent ignorance of their lay brethren.”
Herodotus, and other ancient historians and geographers, have
recorded the migration of a vast body of discontented native soldiers
from Egypt, in the time of Psammeticus. These, we are told, to the
number of 240,000, retired to the country of Ethiopia, where they
were kindly received by the Emperor. They assisted him in his wars,
and in return were apportioned, as a residence, some country on the
confines of Ethiopia, from which they were to drive a rebellious
people to make room for themselves. Herodotus places this country
“upon the Nile, at about the same distance beyond Meroë as this last
is from Elephantine, or fifty days’ journey;” and he also adds, that
“the Antomali (deserters) are known by the name of Asmach, which,
being translated, signifies ‘standing on the left hand’ of the King.” It is
a most remarkable circumstance that the reason or origin of the
name of the country of Gurague, literally “on the left side,” has long
been a question of interest with every Abyssinian traveller, but none
have given any satisfactory explanation for what reason this
particular, and evidently significant, name was first applied. The
situation of the Amhara with respect to the Abi or Bruce’s Nile at
once accounts for the designation, as they live upon the left hand of
the stream as it flows south from lake Dembea, whilst that portion of
this people still retaining their ancient name and purity of descent,
the present Gurague occupy a country similarly situated with respect
to the river Zebee, or Azzabi, or Assabinus, the Ethiopian Jupiter. Abi
and Abiah, other names for branches of the Nile in Abyssinia, are
expressive of father or king, evidently from having been, at a former
period, the chief objects of worship by the people inhabiting their
banks. “Asmach,” and “Gurague,” bear, therefore, the same
interpretation, “to the left of the king,” and none other can explain the
circumstance of the latter name being given to the Amhara. It
appears, however to have been bestowed in contra-distinction to the
“Gongas,” or “Kongue,” a people who originally occupied the right
banks solely of the Abi and Abiah.
This singular correspondence between “Asmach” of the Grecian
historian, and “Gurague” of modern travellers, would be alone,
perhaps, inconclusive evidence that these terms apply to the same
people or country, but some additional evidence may be drawn from
the account which Pliny gives of these Egyptian fugitives. On the
authority of Aristocreon, he states, that “Seventeen days from Meroë
is Esar, a city of those Egyptians who fled from Psammeticus, and
entered the service of the monarch of that country, and in return
received a considerable tract of territory upon the confines, from
which the Ethiopian prince ordered them to expel a tribe of people, at
that time in rebellion against him, and this migration of the Egyptian
troops, introducing the arts and manners of a refined nation, had a
very sensible effect in civilizing the Ethiopians.” The most interesting
particulars we gather from this information, is the name of the city, or,
as I presume, the chief seat of these fugitives, Esar.
By a singular coincidence in the Old Testament, we are told that
Esau is Edom, and although I am not going to infer from this alone,
any connexion between that patriarch and the Ethiopian city, Esar,
yet the philological analogy between the scriptural proper names,
curiously enough, also exists between those of profane history; for
the Esar and Amhara of our subject, express the very same idea as
Esau and Edom, which by all Biblical commentators, is allowed to be
the colour red. “And the first came out red, all over like an hairy
garment; and they called his name Esau.” (Genesis xxv. 25.)
In the present Dankalli language, and I think also, in that of
ancient Meroë, Assar signifies red. In the Persian, I am given to
understand that the planet Mars is called Azer, from its characteristic
colour, a circumstance of significant import when it is considered that
the word Calla, from which is derived Galla, “Ab” the root of Abi, and
“Nil,” from which comes Nile, with others I have yet to speak of, as
designations of places in Abyssinia, are all referable to the same
language. To return, however, to Esar, and its connexion with the
colour red, for it is the same with Esau, and that it is the same as
Edom in Hebrew, I advance the testimony of Dr. Stukeley, who,
speaking of the Red Sea, remarks, “That sea had its name from
Erythras, as the Greeks and the same Pliny write; who is Edom, or
Esau, brother of Jacob. The words are synonymous, signifying
red.”[1] Amhara, also bears the same interpretation in Amharic, and
although it has another meaning, that of beautiful, this is only
because of the national taste directing the name of the favourite
complexion among them, to be employed as the term for beauty
itself. The Dankalli slave-merchant well understands this, for a light-
red Abyssinian girl is the Circassian of oriental harems. In Arabia,
where the original word still conveys the more common idea, we find
“hamah” employed to express the colour red.
In this manner, I connect the “Asmach” of Herodotus, with
Gurague of modern travellers, and the Esar of Pliny, with the Amhara
of the present day, and from these two mutually corroborating
correspondencies, the identity of the modern Abyssinians of Dr.
Prichard with the Automali of Herodotus may perhaps be deduced,
and the difficulty of accounting for a Hebrew people, situated on the
Abyssinian plateau only requires proof to be advanced that the
revolted soldiers of Psammeticus were of the same family of man as
the fugitive Israelites who sought a refuge, under nearly similar
circumstances, in Syria, and built the walls of Jerusalem; and as
their languages are nearly the same, as also their manners,
customs, and ancient religion, previous to the introduction of
Christianity, it will not, perhaps, be difficult to adduce such evidence.
For my part, I am inclined to believe in this national relationship,
because it is partly confirmed by the received account of the
brothers, Esau and Jacob, contained in the book of Genesis, and the
connexion between the two patriarchs, and the country of Egypt, will
perhaps receive some illustration from the opinion I have ventured to
advance upon the subject. In the elder brother, Esau, I perceive the
father of the royal shepherds, and among the list of the dukes, his
descendants may be found, perhaps the pastor kings who held for
some time the sovereign power in Egypt.
The connexion also of the name Esau, or Esar, with the
profession of soldiers, is evident, for in oriental mythology it is
identified with the god Mars; whilst on the other hand, the word
Israel, in Hebrew, I believe, as in Amharic has an immediate
reference to labour, as the name Jacob has also to the heel, which
coincides very singularly with the idea prevalent in India, that the
labouring class have all sprung from the foot of Brahmah. It would be
very interesting, if future discoveries in hieroglyphics, or other
cotemporary histories, which, I believe, do exist in central Africa,
should prove that the appearance of the Jews as a family of man,
under the patriarch Abraham, marks the disruption of an African
community of castes, where the Priest class, excited by the ambition
of a Psammeticus, should determine upon the expulsion of the
soldiers, who thereupon fled to Ethiopia; and, also, that after a
tyrannical and cruel oppression should ultimately occasion the flight
of the workmen, or Israelites, into Palestine. I leave the question,
however, now, to more profound ethnologists, and shall conclude
this, I am afraid, very uninteresting subject, with a short but
necessary description of the features and physical characteristics of
the present Amhara population of Abyssinia.
In the British Museum are many Egyptian statues that possess
exactly the features of the genuine Amhara race. One more,
especially of a woman in the lower saloon marked 16, I will
particularize, to enable those who have the opportunity of examining
these relics of an extinct nation to form a proper idea of the
physiognomy of the people I am speaking of.
Their general complexion cannot be better described by reference
to a familiar object than comparing it with that of red unpolished
copper. Their skin is soft and delicate; the general stature is below
the middle height of Europeans. Their forms are not fully developed
until they have arrived at the same years of puberty as ourselves;
and it is very uncommon for women under seventeen to bear
children. The features of the women conform to a general
characteristic type, and less variations from this are observed among
them than in the men. This observation extends to other races
besides the Amhara, for I have invariably found more consistency of
countenance, more nationality preserved in the features of females
than in the males of the many different people I have met with in my
travels in Abyssinia.
The Amhara face is ovate, having a considerably greater
expression of breadth in the upper than in the lower part. The scalp
in front encroaches upon the forehead, making its length
disproportionate to its height, and, in consequence, it appears
exceedingly low. The eyes are long, but rather full, and the
separation of the eyelids longitudinal, as in Europeans. Their cheeks
are high, yet finely rounded, and sometimes, with the long forehead,
giving to the countenance a nearly triangular form. The nose straight
and well-formed, with a small and beautiful mouth, a finely-curved
edge gradually rising from the commissure to the fulness of a most
inviting pair of lips. A voluptuous fulness, in fact, pervades the whole
countenance; a something more than muscular fibre, yet not exactly
fat, giving a healthy fleshiness, that reminds you of the chubbiness of
children; and I expect the fascinating expression so generally
ascribed to Abyssinian beauties by all orientals is owing to the idea
of innocence and simplicity, that inseparably connects itself with this
infantile character of face. The hair is soft and long; it is neither
woolly, like the negro, nor is it the strong, coarse, straight hair of the
Gongas, or yellow inhabitants of the right bank of the Abi and Abiah
branches of the Azzabi, or red Nile.
I saw few or no cases of distortion among the families I met with
in Efat, and my impression is that they but rarely occur, the natural
and simple lives of the people conducing to easy parturition and a
healthy offspring. The Amhara, however, in their most unchanged
condition in Gurague, and the neighbouring Christian states, have
yet to be visited. The inhabitants of these countries may exhibit
characteristic traits that I have had no opportunities of observing, for
those I met with were the most favourable specimens of the imported
slaves, or their immediate descendants, who were married to
Mahomedans of Efat.
Individuals possessing what I believe to have been the
characteristic features of the genuine Amharic countenance are but
seldom seen on the high land of Shoa, although it might naturally be
expected that their situation would favour a lighter complexion than
the dark-brown Shoans exhibit. This is to be attributed to the very
recent period that their Galla ancestorial relations intruded
themselves into this former Amhara district, as Abyssinian history
records that the first appearance of these invaders from the low
plains of Adal occurred no later than the year 1537.
 From the 27th to the 31st of May, Mr. Scott and I remained in easy
durance at Farree. We were frequently summoned to the presence
of the Wallasmah, whom we would amuse by firing off my gun, or
teaching his son, a boy about fourteen years old, to let off percussion
caps without shutting his eyes. The dreadful experiment would never
be attempted by papa, but he wonderfully enjoyed the bright promise
of his hopeful progeny, the child of his old age, who, on the other
hand, annoyed us not a little by the unsatisfied pertness with which
he demanded to be so indulged.
Day after day were we most solemnly promised that we should
start upon the morrow, but without any intention of being permitted to
do so, beyond the accident occurring of our being sent for by the
King. Perhaps our importunity excited a desire to gratify us, and what
they wished for our sake the kind-hearted people of Farree asserted
would be, because of the great probability that the messenger who
had been sent to the King to receive his commands, would return
sooner than he did.
I am not going to acquit the Wallasmah on this plea, for his want
of courtesy towards us; for from some incomprehensible antipathy,
he would, had he dared, have placed us in irons, and even on
occasions of our visiting him, when we endeavoured to do everything
we could to please him, a surly smile was our only return for some
little gratification we might afford to his boy. His people frequently
made excuses for the conduct of their chief, by stating that he either
had been drinking, or else that he had not; so, drunk or sober, it
seemed quite natural to them that the old fellow should be in a
continual ill-humour from some undefined connexion with strong
drink.
I took care to promise him another present on the occasion of our
leaving Farree, as I conceived that it might be some expectation of
the sort that was operating to cause our tiresome detention. I was
wrong in this, for it was not his pleasure, but the King’s, his master,
that we should be kept at Farree, although he tried to make us
believe it was his own, and assuming an authority that did not belong
to him, made our confinement more irksome than it needed to have
been, on purpose to evince his power. With our sentinels behind us,
however, we could wander all over the hill of Farree, and we
accordingly amused ourselves by endeavouring to extend our
information upon the various subjects of novel interest with which we
were surrounded.
One observation I cannot do better than to insert here, respecting
the rocks and soil of Farree, which abound with the nitrate of potass,
the bald face of the former, in many places, being hollowed into deep
grooves by the constant attrition of the tongues of the numerous
flocks and herds, which seem to be as fond of this salt as the same
animals are of common table salt in other countries; a circumstance
that is well shown in those saline resorts of deer and buffaloes,
called the “licks” of North America. The geological structure of the
hills in this neighbourhood is a finely-grained trachytic rock; grey,
save where the intrusion of narrow dykes of some blacker rocks, a
few feet in thickness, and evidently heated on their first appearance,
has changed the general colour to a deep red, which gradually
recovers its natural hue at the distance of some yards on either side
the dyke. This rock contains a considerable quantity of decomposing
felt-spar, supplying the potass, and, I presume, deriving from the
atmosphere, and the moisture it contains, the other necessary
elements to form the thick efflorescence of saltpetre that covers in
some places the surface of the rock.
The religion of Farree is exclusively Mahomedan, as is also that
of more than three-fourths of the towns and villages of the province
of Efat, all of which are under the hereditary viceregal Wallasmah,
who boasts a descent from the famous Mahomed Grahnè, the Adal
conqueror of many portions of the ancient Abyssinian empire, in the
sixteenth century. Efat forms a portion of the valley country, or
Argobbah, which extends from the edge of the table land of Shoa to
the Hawash, that flows along the base of this slope, from the south
towards the north. The northern boundary of Efat is the river Robee,
the southern one being the Kabani; both of them flow into the
Hawash.
Late in the afternoon of the 30th of May, the messenger returned
from Angolahlah, with orders from the King that I should be allowed
to proceed thither, and that the stores should be conveyed to his