Document 4eg1ZevYL8qzyYjZrnVjdo5a

CHARLES H . HENNEKENS, M.D., D1.P.H. Associate Professor. Departments of Medicine a n d Preventive Medicine and Clinical Epidemiology, Harvard Medial School. Brigham and Women's Hospital: Adjuna Associate Professor of Epidemiology and Biostatistics, Boston University School of Public Health. Boston JULIE E. BURING, SC.D. Assistant Professor. Department of Preventive Medicine and Clinical Epidemiology. Harvard Medical School. Brigham and Women's Hospilal: Adjunct Assistant Profissor of Epidemiology and Biostatistics. Boston University School of Public Health. Boston EDITED BY SHERRY L. MAYRENT, Ph.D. Associate in Medicine (Epidemiology), Harvard Medical School. Brigham and Women's Hospital, Boston FOREWORD BY SIR RICHARD DOLL., F.R.S. Regis Professor Emeritus of Medicine, Oxford University. Oxford, England LITTLE, BROWN AND COMPANY BOSTONITORONTO- c Copvright 8 1987 by Charles H. Hennekens andJulie E. Buring First Edition Third Prrnrrng All rights reserved. No part of this book rnav be reproduced in any form or by anv electronic or mechanical means, including information storage and retrieval systems. without permission in writing from the publisher. except by P reviewer who may quote brief passages in a review. Library of Congress Catalog Card No. 87-45377 ISBN 0-516-35636-0 Printed in the United States of America \ Contents Foreword xi Preface xiii Acknowledgments sv I. BASIC CONCEPTS 1. Definition and Background 3 Hisiorv of Epidemiolocy 4 Chanpng Patterns o f Mlortalitv 8 Drvelopinents in hioderii Epidernioloy\. IO T h e Unique Contribution of Epidemiology l ? 2. Design Strategies in Epidemiologic Kesearch Descriptive Studies I6 Aiialytic Epitieniiology 2 I 16 3. Statistical Association and Cause-Effect Relationships Evaluation of the Presence of a Valid Statistical Association Judgment of a Cause-Effect Relationship 39 T h e Evaluation of a Hypothesis: Smoking and Lung Cancer 30 31 44 4. Measures of Disease Frequency and Association Measures or Disease Frequency 54 PREVALENCE AND INCIDENCE 57 CRUDE. CATEGORY-SPECIFIC. AND ADJUSTED (ST.4NDARDIZED) RATES 66 Measures of Association 73 PRESENTATION OF DATA 73 RELATIVE RISK 77 STANDARDIZED LIORTALITY RATIOS 82 ATTRIBUTABLE RISK 87 -INTERPRETATION OF MEASURES OF ASSOCIATION 54 9s vii viii CONTENTS i II. TYPES OF EPIDEMIOLOGIC STUDIES 5. Descriptive Studies 101 Types of Descriptive Studies 101 CORRELATIONAL STUDIES 102 CASE REPORTS AND CASE SERIES 106 CROSS-SECTIONAL SURVEYS 108 Hypothesis Formulation from Descriptive Studies 112 6. Case-Control Studies 132 Issues in the Design and Conduct of Case-Control Studies lssues in Analysis 144 Issues in lnterpreution 145 134 7. Cohort Studies 153 Tvpes of Cohort Studies 154 Issues in the Design and Conduct of Cohort Studies lssues in Analysis 168 lssues in Interpretation 168 156 8. Intervention Studies 178 Tvpes of Intervention Studies 179 Unique Problems of Intervention Studies 180 Issues in the Design ant1 Conduct of Clinical Trials 181 Stopping Rules: Decision for Early Tern\ination of a Trial Sample Size Considerations: Statistical Power 198 Issues in Analysis and Interpretation 204 196 III. DESCRIPTION AND ANALYSIS OF EPIDEMIOLOGIC DATA 9. Presentation and Summarization of Data 215 Tyhs of Variables 215 Dam Presentation ?18 Summary Statistics 229 MEASURES OF CENTRAL TENDENCY 231 MEASURES OF SPREAD OR VARIABILITY 235 10. Analysis of Epidemiologic Studies: Evaluating the Role of Chance 243 Inference 243 Hvpothesis Testing 244 Estimation: Confidence Intervals 252 Study Design Issuer: Sample Size and Power Statistical Overviews 264 258 i. CONTENTS ix 1 1. Analysis of Epidemiologic Studies: Evaluating the Role of Bias 272 Types of Bias ?7? Control of Bias: Studv Design 276 Evaluation of the Role of Bias 282 12. .4nalysis of Epidemiologic Studies: Evaluating the Role of Confounding 287 T h e Nature of Confounding 287 Methods to Control Confounding in the Design Methods to Control Confounding in the Anaiysis ?95 304 IV. EPIDEMIOLOGY IN DISErfSE CONTROL 13. Screening 327 Diseases Apprupriare for Screening 328 Screening Tests 331 Evaluarion of Screening Programs: Feasibility and Efficacy 335 APPEiVDIXES A. Standard Scatistical TJbles 35 1 B. Selection of an Appropriate Tesr of Statistical Significance 356 C. Answers to Study Qucstions 360 Index 373 100. I`liose conLuming illy Iiigher C H D more that of nondrinkcrs ters. Such a nonliriear ional studies, in which ion for a population le individuals. ' tions, case reports and ierit or a group of palies, in which typically a disease or a patient's iesis. I n this way, case el-facebetween clinical :s of studies published of.d l articles in one nusual medical occurmtification of new dise , in recent years, case !ards related to a nurnrding "Frisbee finger" rist" [44], and "breakapter 2, it was a single ypotliesis that OC use wrts, which mav occur I has historical impor?atI) tileans to identify at present, the routine ~ g g e s t sthe emergence of the activities of the ad to formulation of a requeritly using a casesperiences of the case not develop the disease liance on a single case ing a useful hypothesis Iddity. For example, in :ries of three men with 1y1 chloride plant. This 5 . DESCRIPTIVE STUDIES 107 i i t t i n I ) ( ~ rit1 sucli ;i sni;i11 popu1;irion during the time period studied was c-learly i n cscess of w h a t w;is expected and led to the formulation of the tlvpotliesis tliat occupatiori;il exposure to vinyl chloride caused hepatic ~ ~ i ~ ~ i o s ~ i r cLoat er ri ~thi .e same year, this hypothesis was substantiated by data from t w o analytic stuciies [57, 691. A single case report from that plant might not have been sufficient to allow formulation and subse- quent testing of this hypothesis. T h e usefulness of case reports and case series in the recognition of iiew diseases and the formulation of hypotheses concerning possible risk [actors can be illustrated by the early epidemiology of acquired immu,iotieficiencv syndrome (AIDS) [61]. Between October 1980 and May 1`3X 1 , 5 cases of Pneumorjwtis carinii pneumonia were reported among \ oung, previously healthy, homosexual men in Los Angeles [lo]. This case series was unusual in that this type of pneumonia had previously occur-red o n l y in older cancer patients whose immune systems were supjlressed, usually as a result of chemotherapy [75]. Similarly, in early 1!)8 I . an unprecedented number of cases of Kaposi's sarcoma were dixgnosed in young homosexual men. Again, this was noteworthy because previousl? this malignancy had been seen almost exclusively in the eltIci-l\ and affected men and women equally [ I l]. As a result of these case \cries, the Centers for Disease Control immediately initiated a surveil- I`il1ce prosrani to quantitate t h e magnitude of this problem and to de\c.lop diagnostic criteria for what appeared to be a new disease. This pi-oqram quickly identified that homosexual men were at high risk of cie\,elopinf:this syndrome. Subsequent case reports and case series sug- ;;csted that ;\IDS also resulted from blood-borne transmission among Io1raveiioiis drug abusers as well as recipients of transfusions [2] and Ilemophiliacs receiving / h o d products [121. These descriptive data prolided clues used in the design and conduct of analytic studies, which suhsequently led to the identification of a number of specific risk factors f o r the cievelopment of. AlDS. T h e availability of sera from these cases a n d cornparable controls also has contributed to the identification of hut m i i immunodeficiency virus ( H I V ) as the transmissible agent [61]. \Vhile case reports and case series are very useful for hypothesis for- mulation, they cannot be used to test for the presence of a valid statistical association. One fundamental limitation of the case report is that it is based on the experience of only one person. T h e presence of any risk .~ factor, however suggestive, may simply be coincidental. Although case s-e--r-i_e_s_.are .f..-r.equIe__n_.t_l._y~ suf.ficiently large-Lo-p-it-quantification offre- (.l.w.-.n--cy .o-.f an exposure, the interpretability of such information is se- yerely limited bv the lack of an appropriate comparison group. This lack can eitlier ohscure a relationship or suggest an association w h e E c o n e actually exists. For exatnple, in a n attempt to explore the possible asso- clatiori tmween the use of OCs and hepatocellular carcinoma in women,