PDQ EPIDEMIOLOGY Third Edition FM.indd i 7/3/09 12:43:38 AM PDQ* SERIES Also in this series: NORMAN, STREINER PDQ STATISTICS SCHLAGENHAUF-LAWLOR, FUNK-BAUMANN PDQ TRAVELER’S MALARIA MC KIBBON, WILCZYNSKI PDQ EVIDENCE-BASED PRINCIPLES AND PRACTICE 2/E *PDQ (P r e t t y D a r n e d Q u i c k ) FM.indd ii 7/3/09 12:43:38 AM PDQ EPIDEMIOLOGY DAVID L STREINER, PhD Senior Scientist, Kunin-Lunenfeld Applied Research Unit Baycrest Professor, Department of Psychiatry University of Toronto Toronto, Ontario GEOFFREY R NORMAN, PhD Canada Research Chair in Cognitive Dimensions of Clinical Expertise Professor of Clinical Epidemiology and Biostatistics Assistant Dean, Programme for Educational Research and Development McMaster University Hamilton, Ontario both of Mc Master University Faculty of Health Sciences Hamilton, Ontario T HIR D E DI T ION with 43 illustrations 2009 People’s Medical Publishing House—USA Shelton • Connecticut FM.indd iii 7/13/09 10:18:55 PM People’s Medical Publishing House—USA Enterprise Drive, Suite 509 Shelton, CT 06484 Tel: 203-402-0646 Fax: 203-402-0854 E-mail: info@pmph-usa.com © 2009 People’s Medical Publishing House—USA, Ltd All rights reserved No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior written permission from the publisher 09 10 11 12/PMPH/6 ISBN 978-1-60795-022-6 Printed in China by People’s Medical Publishing House Sales and Distribution Canada McGraw-Hill Ryerson Education Customer Care 300 Water St Whitby, Ontario L1N 9B6 Canada Tel: 1-800-565-5758 Fax: 1-800-463-5885 www.mcgrawhill.ca Foreign Rights John Scott & Company International Publisher’s Agency P.O Box 878 Kimberton, PA 19442 USA Tel: 610-827-1640 Fax: 610-827-1671 Japan United Publishers Services Limited 1-32-5 Higashi-Shinagawa Shinagawa-ku, Tokyo 140-0002 Japan Tel: 03-5479-7251 Fax: 03-5479-7307 Email: kakimoto@ups.co.jp United Kingdom, Europe, Middle East, Africa McGraw Hill Education Shoppenhangers Road Maidenhead Berkshire, SL6 2QL England Tel: 44-0-1628-502500 Fax: 44-0-1628-635895 www.mcgraw-hill.co.uk Singapore, Thailand, Philippines, Indonesia, Vietnam, Pacific Rim, Korea McGraw-Hill Education 60 Tuas Basin Link Singapore 638775 Tel: 65-6863-1580 Fax: 65-6862-3354 www.mcgraw-hill.com.sg Australia, New Zealand Elsevier Australia Tower 1, 475 Victoria Avenue Chatswood NSW 2067 Australia Tel: 0-9422-8553 Fax: 0-9422-8562 www.elsevier.com.au Brazil Tecmedd Importadora e Distribuidora de Livros Ltda Avenida Maurilio Biagi 2850 City Ribeirao, Rebeirao, Preto SP Brazil CEP: 14021-000 Tel: 0800-992236 Fax: 16-3993-9000 Email: tecmedd@tecmedd.com.br India, Bangladesh, Pakistan, Sri Lanka, Malaysia CBS Publishers 4819/X1 Prahlad Street 24 Ansari Road, Darya, New Delhi-110002 India Tel: 91-11-23266861/67 Fax: 91-11-23266818 Email:cbspubs@vsnl.com People’s Republic of China PMPH Bldg 3, 3rd District Fangqunyuan, Fangzhuang Beijing 100078 P.R China Tel: 8610-67653342 Fax: 8610-67691034 www.pmph.com Notice: The authors and publisher have made every effort to ensure that the patient care recommended herein, including choice of drugs and drug dosages, is in accord with the accepted standard and practice at the time of publication However, because research and regulation constantly change clinical standards, the reader is urged to check the product information sheet included in the package of each drug, which includes recommended doses, warnings, and contraindications Th is is particularly important with new or infrequently used drugs Any treatment regimen, particularly one involving medication, involves inherent risk that must be weighed on a case-by-case basis against the benefits anticipated The reader is cautioned that the purpose of this book is to inform and enlighten; the information contained herein is not intended as, and should not be employed as, a substitute for individual diagnosis and treatment FM.indd iv 7/3/09 12:43:38 AM To two individuals who have provided comfort and solace in the difficult times and added to our pleasure in the good times— Jim Beam and Johnny Walker FM.indd v 7/3/09 12:43:39 AM FM.indd vi 7/3/09 12:43:40 AM Contents Preface, xi Introduction to Epidemiology, What It Is, Trends in Epidemiology, Current Applications of Epidemiology, A Dose of Reality, 13 Classical Epidemiology, 21 A Little Bit of History, 21 Some Basic Concepts, 25 Risk Factors, 32 Some Other Terms You Should Know, 34 Research Methodology, 37 Yet Some More History, 38 Design Elements, 40 Sampling, 46 Subject Allocation, 52 Other Forms of Randomization, 58 Threats to Validity, 63 Research Strategies, 85 Meta-Analysis, 99 C.R.A.P Detectors, 102 vii FM.indd vii 7/3/09 12:43:40 AM viii CONTENTS Measurement, 111 Issues in Choosing a Measure, 112 Types of Variables, 114 Measurement with Categorical Variables, 115 Measurement with Continuous Variables, 145 C.R.A.P Detectors, 153 Assessing Causation, 159 The Criteria, 160 C.R.A.P Detectors, 167 Ethics, 173 Freely Given Consent, 174 Informed Consent, 176 Other Issues, 182 Appendix, 187 A Brief Epidemish-English Dictionary, 187 Index, FM.indd viii 191 7/3/09 12:43:40 AM Preface Welcome to the wonderful world of epidemiology Just when you figured that you had mastered the mysteries of pulmonary blood flow, cardiac rhythms, electrolyte balance, gut motility, and cerebral anatomy, along came this strange guy in a tweed jacket muttering formulas, statistics, and foreign-sounding words like “relative risk,” “positive predictive value,” and “Mantel-Haenzel chi-square.” He didn’t look like a scientist: no dirty lab coat, scientific calculator, or long hair He didn’t look like a real doctor either: no clean lab coat, stethoscope, or designer length hair Yet he had the arrogance to claim that he is both a clinician and a scientist—he is a clinical epidemiologist Amid the hushed silence in the room you can overhear desperate whispers of, “What in the world is a clinical whatever-it-is-ologist?” More particularly, why, in a world already overpopulated by physiologists, pharmacologists, pathologists, gerontologists, nephrologists, cardiologists, neurologists, and a dozen other relatively legitimate art forms, we need yet another -ologist? The answer, it seems, is that somewhere in that complex, compartmentalized world that lays claim to the human body as an object of study, common sense got lost in the shuffle The reality is that, despite a tremendous explosion in biomedical science, we still know embarrassingly little about the workings of ourselves No one knows the cause of most diseases or the cure for that matter No one can lay claim to the crystal ball that will predict accurately who, among a group of cancer or postmyocardial infarction patients, will survive a year As a result, there is a considerable gap between the exact findings of the laboratory and the uncertain world of clinical medicine This leaves enormous room for the dissemination of wellintended but useless tests, therapies, or theories Some examples may illustrate the point: As we have been told by the media on countless occasions, dioxin is about the most lethal chemical known A tiny dose causes mice to curl up and die; dioxins are teratogenic, mutagenic, and carcinogenic in the lab Yet despite all the tons of Agent Orange dumped over Vietnam and ix FM.indd ix 7/3/09 12:43:40 AM x PREFACE some large spills in places like Seveso, Italy, there is no evidence that they are a significant human carcinogen Conversely, cigarette smoking is easily the most lethal human carcinogen, measured in the number of lives lost Yet it was long after good scientific evidence from human studies convinced everyone (except tobacco farmers and cigarette manufacturers) that it caused cancer that scientists were able to induce cancers in mice in the laboratory Clofibrate was a very popular lipid-reducing agent in the mid-1960s There was abundant laboratory evidence that it would work as claimed Unfortunately, later randomized trials proved that the drug killed more people than it saved We like to think that the days of patent medicines and snake oil salesmen have passed (However, one visit to your local drugstore to peruse the over-the-counter antiarthritis drugs, none of which contains anything more than aspirin and all of which cost 10 times as much, should dispel that myth.) Nevertheless, mainstream medicine is still susceptible to the legitimate and honest claims of success of new therapies based on experience with few patients Many of these therapies are eventually proved to have no value One case in point is gastric freezing There were a number of case reports, involving a total of about 1,500 patients, that indicated that it would cure ulcers It was only later that trials demonstrated that the procedure was useless Whatever happened to tonsillectomies? It seems as if five out of six adults over the age of 40 had their tonsils removed in childhood, but very few of our children have to endure this agony Credit for the turnaround belongs to one of the neater epidemiologic studies It was common wisdom in those days that roughly half of all kids needed their tonsils removed These investigators started with about 400 kids who still had their tonsils and shipped them around to a group of respected physicians Sure enough, 45 percent of tonsils had to go The researchers removed these “diseased” kids from the study and sent the remaining ones around again (to different physicians, of course) This time 46 percent of the tonsils were slated to go Now, the kids who were left (who had been judged healthy by two sets of physicians by this time) were marched before a third group of doctors Want to guess how many were said to need tonsillectomies? You got it—44 percent These examples nicely illustrate the role of epidemiology these days—it comfortably fi lls the gaping chasm between the scientific wisdom of the wet laboratory and the clinical wisdom of the ward The good news is that it isn’t all that hard Despite the fancy terminology, epidemiology is, above all, the science of common sense (Its bedmate, biostatistics, isn’t quite so straightforward To decipher the arcane logic of statisticians, we heartily recommend another book in the PDQ series—PDQ Statistics We’re biased, of course, because we wrote it.) FM.indd x 7/3/09 12:43:40 AM 184 PDQ EVIDENCE-BASED PRINCIPLES AND PRACTICE amounts of ASA for their headaches What is more likely, in these days of lookalike drugs, is that there is no appreciable difference Now what are the possible reasons for the lack of difference? The first is that the drugs really are not all that different in anything except price, and we would be correct in staying with the cheaper ASA The next reason is that one drug actually is better than the other, but the trial wasn’t large enough to detect a difference; in essence, a Type II error Last, the drugs may be different and there may have been enough subjects, but the study was badly designed or executed— poorly chosen and insensitive outcome measures, badly trained raters, noncompliant patients, or a host of other factors In this case, because there is no placebo arm with which to compare the results, it is impossible to know if both drugs were equally effective or were equally ineffective in the hands of this research team If the ASA group did not any better than a placebo group, though, it would have signaled to us that the study was flawed, and we shouldn’t trust the results Another reason for a placebo group was seen in a recent trial of flosequinan for treating heart failure When it was compared with an angiotensin converting enzyme (ACE) inhibitor, the two were found to be equally efficacious, which would indicate that it is an acceptable substitute, especially if it were cheaper When either flosequinan or placebo were given to patients receiving an established therapy, though, a different story emerged; it significantly increased the risk of death A third reason in favor of comparing a new treatment against a placebo rather than an alternative is that, paradoxically, fewer people are put at risk of adverse drug reactions (ADRs) Let’s assume that ADRs occur 10% of the time with both drugs When we compare an active drug against a placebo, then we usually expect a fairly large difference, so the sample size needs to be relatively small; say, 50 patients per group This means that nobody in the placebo group will have an ADR, and five patients in the new treatment group will Now, if we compare the new treatment against an established one, the difference between the two will be much smaller, so that a far larger sample size is needed; say, 100 per group In this case, 10 patients in each group will experience ADRs, four times the number in a placebo-controlled trial Thus, it may actually be more ethical to use a placebo comparison rather than an active control group A final objection to the placebo group is that it always deprives half the patients of receiving an effective therapy If the new drug is compared with standard therapy and the new drug works, then all patients receive some benefit; if it doesn’t work, at least half have gotten something effective In a placebo trial, if the treatment works, half benefit; if it doesn’t work, nobody gets anything effective With this logic, it would always be better to omit the placebo arm However, this assumes that all treatments are safe As we saw with flosequinan and many other drugs, side effects exist and Chapter 6.indd 184 6/13/09 5:35:49 PM Chapter Ethics 185 can sometimes be fatal, even for accepted treatments Those in the placebo group may not enjoy the benefits of treatment, but by the same token, they are not subjected to the risks So what’s the bottom line? If the disease is life-threatening or may result in irreversible injury and a proven therapy exists, there is no issue; a placebo group would be patently unethical At the opposite extreme, if the disease is relatively benign, and the placebo patients are simply delayed in receiving a proven therapy for a few days or weeks, there are likely few ethical concerns with using a placebo group The problem, as always, lies in the middle Here the balance is between the risk of allowing a disease to remain untreated for a given length of time, versus the benefit of measuring the exact effect of the treatment, as opposed to its effect relative to some other intervention and having more confidence in results showing equivalence As the risk and the interval increase, the balance is tipped toward using only an active comparison; as the risk and duration of the trial decrease, the direction is toward a placebo arm The result of this calculus is highly subjective and dependent on the local IRB REFERENCES American Psychological Association: Ethical principles of psychologists and code of conduct, Am Psychol 47:1597–1611, 1992 Cowley AJ, Wynne RD, Swami A: A comparison of the effects of captopril and flosequinan in patients with severe heart failure, Cardiovasc Drugs Ther 6:465–470, 1992 Hill AB: Medical ethics and controlled trials, Br Med J 1:1043–1049, 1963 Medical Research Council of Canada: Guidelines on research involving human subjects, Ottawa, 1987, MRC NCBHR Office: Divided loyalties: an anthology of conflict of interest duties, NCBHR Communiqué 6:11–16, 1995 Office for Protection from Research Risks: Exempt research and research that may undergo expedited review, OPRR Reports, No 95–02, Rockville MD, National Institutes of Health, 1995 Office for Protection from Research Risks: Protection of human subjects, OPRR Reports, Code of Federal Regulation 45 CFR 46, Rockville MD, National Institutes of Health, 1983 Packer M, Rouleau J, Swedberg K: Effect of flosequinan on survival in chronic heart failure: preliminary results of the PROFILE study, Circulation 88:(suppl I):301, 1993 (Abstract) Rothman KJ, Michels KB: The continuing unethical use of placebo controls, N Engl J Med 331:394–398, 1994 Streincr DL: The ethics of placebo-controlled trials, Can J Psychiatry 40:165–166, 1995 (editorial) Streiner DL: The lesser of two evils: the ethics of placebo-controlled trials Can J Psychiatry 53:430–432, 2008 Chapter 6.indd 185 6/13/09 5:35:49 PM 186 PDQ EVIDENCE-BASED PRINCIPLES AND PRACTICE TO READ FURTHER Beauchamp TL, Childress JF: Principles of biomedical ethics, ed 3, New York, 1989, Oxford University Press Klerman GL: Scientific and ethical considerations in the use of placebo controls in clinical trials in psychopharmacology, Psychopharcacol Bull 22:25–29, 1986 Levine R: Ethics and regulations of clinical research, ed 2, Baltimore, 1986, Urban and Schwarzenber Chapter 6.indd 186 6/13/09 5:35:50 PM Appendix A BRIEF EPIDEMISH-ENGLISH DICTIONARY In the course of writing their reports and journal articles, researchers in epidemiology often use words or phrases whose meanings are somewhat obscure To assist the reader in understanding these terms (and to provide a little amusement), we provide herein a brief dictionary To begin, we offer the definition of clinical epidemiology (itself an obscure term), which is credited to Dr Stephen Leader of the University of Sydney: “Clinical epidemiology is that branch of alchemy whose goal it is to turn bulls—t into airline tickets.” And now to the dictionary When The Researcher Says He or She Really Means A trend was noted The statistical test was not significant The demographic characteristics of the nonresponders were similar to those of the rest of the sample All we really had on them were age and sex Agreement between the raters was acceptable The agreement was so bad that we don’t dare to include the actual number in the paper The questionnaire was circulated to a panel of experts to establish face validity Our friends liked it and the bottle of scotch we included The rate of lung cancer among the hourly rate employees was significantly higher, which may be caused by excess PCBP exposure It might also be caused by obvious things like smoking and social class, but I’m interested in PCBP today 187 Appendix.indd 187 6/13/09 5:19:16 PM 188 PDQ EVIDENCE-BASED PRINCIPLES AND PRACTICE In a case series of 12 patients, showed clinically significant improvement on the experimental drug With the help of the drug company representative, I judged which patients got better under my care The correlation was highly significant (p < 0.0001) With 10,000 subjects, any correlation is highly significant The response rate was 60%, which is acceptable for studies of this type However, the study itself was so bad that even a 100% response rate wouldn’t have saved it Although there was no overall difference in mortality, the rate of left clavicular cancer was higher in blueeyed women in the exposed group If you look at enough things, sooner or later one of them is bound to turn out to be significantly different Although the results appear to be consistent with the predictions, further research is warranted I’ve already applied for a new grant this year Further research is required to clarify the results I haven’t a clue what it all means The difference was statistically significant (p < 0.0001) but clinically useless The study was a single-blind trial Everybody knew who was getting what except the poor patient A retrospective study was conducted We had all these data sitting around and needed some fast publications Morbidity and mortality from Streinorman’s disease represents a significant burden on society It’s my own narrow interest, but I have to justify the research somehow The overall agreement was 87%, which represents a truly remarkable rate of agreement (κ = 0.12) Chance-corrected agreement was so abysmal that we thought we had better talk about raw agreement Based on current trends, the incidence of self-pollution in the year 2010 will be Draw a straight line through the data from one hospital in 1990 and 2000, and that’s what we got It is widely known that I can’t be bothered to look up the reference A one-tailed test was used The results wouldn’t be significant with a two-tailed test After adjusting for baseline differences between the groups We did a lousy job of randomizing Appendix.indd 188 6/13/09 5:19:24 PM 189 Appenidx After adjusting for confounders Boy, did these groups differ! One possible explanation for these results is I can only think of one Forty patients agreed to participate The others were able to pay their hospital bills After conducting a pilot study, we decided to use a mailed questionnaire We got tired of people slamming the phone in our ear After conducting a pilot study, we decided to use face-to-face interviews They wouldn’t return the mailed questionnaires either The data were normalized by truncating outliers We couldn’t get the results we wanted, so we threw out subjects until we got what we were looking for We did not include premorbid status and number of previous hospitalizations in the model We forgot to gather these data The agreement between raters was: excellent good acceptable low The agreement between raters was: good fair nonexistent negative Data were analyzed using the Schmedlap-Scheisskopf test We tried the usual tests, but they didn’t give significant results A fiducial reference line A scratch Although it has not been possible to provide definite answers to these questions The experiment didn’t work out, but I figured I could at least get a publication out of it Three of the samples were chosen for detailed study None of the other samples worked, so we threw them out Correct within an order of magnitude Wrong The assays were handled with extreme care during the experiments We didn’t drop any on the floor No one has looked at this before Nobody else really cares about this Time and time again I’ve seen two cases In case after case after case I’ve seen three cases Appendix.indd 189 6/13/09 5:19:28 PM Appendix.indd 190 6/13/09 5:19:28 PM Index A Absolute risk reduction, 129 “Accidents,” nuclear, 7–8 Accuracy, 141, 142 Acquired host factors, 27 Acquired immunodeficiency syndrome, 132 Active surveillance, Adaptive sampling, 59 Admission of psychiatric patients, involuntary, 41 Adults, incompetent, informed consent from, 179 Age and changes in inductive reasoning, 84 Age of Degenerative and Man-made Diseases, Age of Receding Pandemics, Age of Useful Interventions, Age-specific mortality, 122–123 Agents, 26–27 AIDS; see Acquired immunodeficiency syndrome Allergens, 26 Allocation haphazard, 58 nonrandom, 58 randomized, 52–54 problems with, 58–59 stratified, 55 subject nomenclature for, 86 in research methodology, 52–58 Ambulatory mental health care, Medicaid coverage and use of, 12 Analogy, 166 Analysis, policy, 12 Analytic designs for research studies, 87–91 Anemia, sickle cell, 27 Annual incidence, 117, 118 Annual mortality, 120, 121 AR; see Attributable risk ARR; see Absolute risk reduction Association consistency of, 160–161 measurement of, 145–149 with categoric variables, 126–131 with continuous variables, 145–149 specificity of, 161–162 strength of, 160 between variables, 126 temporality of, 162–163 Attention control group, 70 Attributable risk, 129 Autonomy of individual, 173 Availability bias, 74 B Bacteria, 27 Bayes’ theorem, 138 Beck Depression Inventory, 151 Behavioral host factors, 29 Beriberi, thiamine deficiency and, 26 Berkson’s bias, 67–68, 152 Beta-Blocker Heart Attack Trial, 69 Bias(es), 152 availability, 74 Berkson’s, 67–68, 152 diagnostic suspicion, 74–75 healthy workers, 65–66 incidence-prevalence (Neyman), 66–67 measurement, 152–153 Neyman, 66–67, 152 non-random sampling as form of, 65 social desirability, 153 subject selection, 65–70 volunteer, 68–69 Bifocals, need for, and nocturnal enuresis, relationship between, 76 Biologic gradient, 163–164 Biologic plausibility, 165 Birth cohort, 27, 47, 81–83 Black death, 25, 32 Bleeding, vaginal, and endometrial cancer, association between, 67–68 Blinding, 53, 70–71 Block randomization, 54–55 Breast cancer, 10 dietary fat intake and, 30 Bubonic plague, 25, 34 Burlington Randomized Trial, 50 C Calcium kidney stones and, 26 osteoporosis and, 26 Canadian cancer statistics, 121 Cancer breast, 10 dietary fat intake and, 30 Canadian statistics on, 121 191 Index.indd 191 6/30/09 10:52:11 PM 192 PDQ EVIDENCEBASED PRINCIPLES AND PRACTICE cholesterol level and, 162–163 endometrial, and vaginal bleeding, association between, 67–68 gastrointestinal tract, 30–31 lung, smoking and, 130 proportional mortalities for, 122 respiratory, versus cancers from all other sites, 126 standardized mortality per 1000 for, 125 Caries, fluoridation and, 31 Carriers, 34 Case control studies, 6, 91 Case fatality rate, 119–120 Categoric variables, 114–115 measurement with, 115–144 measures of association with, 126–131 Causality, criteria for, 166 Causation, assessing, 159–166 Cause of new syndrome, identifying, 4–7 Center for Epidemiological Studies– Depression scale, 151 Chance correction using Cohen’s kappa, 141–143 Characteristic curves, receiver operating, 139–140 Chemical agents, 26 Chernobyl, 8, Children, informed consent and, 178–179 Cholera, historical study of, 23–25 Cholesterol, 26 cancer and, 162–163 Cholestyramine study, 127 Classical epidemiology, 2, 16, 21–34 basic concepts of, 25–32 history of, 21–25 Clinical epidemiology, 2–3, 10, 16, 39, 145 Cluster sampling, 50–51 Coercion of research subjects, 174–175 Cohen’s kappa, chance correction using, 141–143 Coherence, 165 Cohort, 27, 47, 81–83 Cohort effects, 81–83 Cohort study, 86, 90–91 Cointervention, 79–80 Comparison groups, 43–45 Compliance bias, 69, 144 Concurrent control group, 44 Concurrent validity, 151 Confounding, 75–77 interactions versus, 78–79 Confounding factors, 25 Consent, 174 freely given, 174–176 informed; see Informed consent no need for, 181–182 Index.indd 192 verbal, 180 Consistency of association, 160–161 Construct validity, 151–152 Contamination, Content validity, 150 Continuous variables, 114–115 measurement with, 145–153 measures of association with, 145–149 Control group attention, 70 concurrent, 44 historical, 44 matching in, 57 undermatched, 57 Controlled trial, randomized, 10, 15, 39, 61, 97, 180 Controls, placebo, ethics of, 183–185 Convenience, sample of, 51 Coronary Drug Project, 69 Coronary heart disease, cholesterol and, 26 Coronary Primary Prevention Trial, 126 Correction, chance, using Cohen’s kappa, 141–143 Correlation intraclass, 148–149 Pearson product-moment, 146 Covariate, 77 Criterion validity, 151 Cross-over design, 98 Cross-sectional survey, 87 Cultural background, informed consent and, 178 Cut-point for test for myocardial infarction, 139 D Ds, measurement of illness using, 112–114 Data gathering, direction of, 42–43 Daydreaming, 27, 28 Declarations of Helsinki, 173, 183 Dependent variable, 40, 72 Depression measures of, 151 prediction of, 141 Descriptive designs for research studies, 87–91 Descriptive studies, 1–2 Design elements in research methodology, 40–46 Desirability bias, social, 153 Diagnostic suspicion bias, 74–75 Diagnostic tests, 131–136, 142, 143 Dietary fat intake and breast cancer, 30 Directionality of events, determining, 42 6/30/09 10:52:12 PM 193 Index Disease(s) endemic, 34 heart, coronary, cholesterol and, 2, 26, 126 Hodgkin’s, incidence of, by age, 29 infectious historical studies of, 23 outbreaks of, Tay-Sachs, 27 Dose-response relationship, 163–164 Double blind study, 70 Duration, 55, 119 E Ecologic fallacy, 84–85 Ecologic study, 89 Economics, health, 12 EF; see Etiologic fraction Effectiveness of treatment, determining, 9–11 ELISA; see Enzyme-linked immunosorbent assay (ELISA) test Emphysema, 27 End points, surrogate, 72–73 Endemic disease, 34 Endometrial cancer and vaginal bleeding, association between, 67–68 Enuresis, nocturnal, and need for bifocals, relationship between, 75–78 Environment, 30 Enzyme-linked immunosorbent assay (ELISA) test, 132 Epidemic, 1, 3, 22, 34 Epidemiologic research, stages of, 1–2 Epidemiologic transition, Epidemiology big-E; see Classical epidemiology classical, 2, 16, 21–34; see also Classical epidemiology clinical, 2–3, 10, 16, 39, 145 current applications of, 4–12 defi nition of, 1–3 introduction to, 1–16 research strategies for, 85–98 research studies for, nomenclature for, 86 trends in, 3–4 Error, random, 149, 153 Erythrocyte sedimentation rate, active joints and, association between, 145–146, 147 Ethical issues in measurement, 113 Ethics, 173–185 methodology and, 182 of placebo controls, 183–185 Etiologic fraction, 128 Index.indd 193 Evidence, experimental, 165–166 Experimental designs for research studies, 92, 97–98 Experimental evidence, 165–166 Experimental studies, 40, 52 Explanatory variables, surrogate, 73 Exposure assessing risks associated with, 7–9 to radiation, risks of, 8–9 F Face validity, 150 Fallacy, ecologic, 84–85 False negative rate, 132–134 False positive rate, 132–134 Fat intake and breast cancer, 2, 30 Fatality rate, case, 119–120 Flu, cyclical pattern of, 32 Fluoride, caries and, 31 Fraction, etiologic, 128–129 Freely given consent, 174 Frequency match, 77 Frequency, measures of, 116–120 G Gastrointestinal tract, cancer of, 30–31 Generalizability of research results, 53 Gradient, biologic, 163–164 Graunt, John, 22 Groups, special, informed consent by, 178–180 H Hantavirus outbreak in midwestern United States, Haphazard allocation, 58 Haphazard sampling, 51–52 Hawthorne effect, 69–70 Headaches, placebo effect in treatment of, 70 Health care, mental, ambulatory, Medicaid coverage and use of, 12 Health economics, 12 Health service, use needs and trends in, identifying, 11–12 Healthy workers bias, 65–66 Heart attack, oral contraceptive use and smoking in risk for, 78 Heart disease, coronary, cholesterol and, 2, 26 Hemophilia, 27 Hippocratic writings, 22 Hiroshima, 6/30/09 10:52:12 PM 194 PDQ EVIDENCEBASED PRINCIPLES AND PRACTICE Historical control group, 44 HIV antibodies, prevalence of, 132, 136 Hodgkin’s disease, incidence of, by age, 29 Host factors, 27–29 Humors, imbalance of, disease and, 22 Hypertension, salt and, 26 Hypothesis testing, I ICC; see Intraclass correlation Impact, measures of, 120–126 Incidence, 117–118 annual, 117, 118 prevalence and, relation between, 119 Incidence-prevalence bias, 66, 87, 103 Incompetent adult patients, informed consent from, 179 Incubation period, 23 Independent variable, 40, 72 Individual, autonomy of, 173 Individual matching, 57 Inductive reasoning, changes in, with age, 84 Infarction, myocardial cut-point for test for, 139 oral contraceptive use and smoking in risk for, 78 Infectious agents, 26 Infectious diseases historical studies of, 23 outbreaks of, Informed consent, 174, 176–182 children and, 178–179 cultural background and, 178, 180 elements of, 176–178 from incompetent adult patients, 179 language of, 178 by special groups, 178–180 Intelligence quotient, decline in, 83 Interactions, 78–79 Interobserver reliability, 149 Interval variables, 115 Intervention, nomenclature for, 86 Intervention studies, Intraclass correlation, 148–149 Involuntary admission of psychiatric patients, 41 Ionizing radiation, 26 J Joints, active, erythrocyte sedimentation rate and, association between, 145–146, 147 Index.indd 194 K Kappa, 142 Cohen’s, chance correction using, 141–143 Kidney stones, calcium and, 26 Koch’s postulates, 159 Kwashiorkor, protein deficiency and, 26 L Language of consent, 178 Legionnaires’ disease, Leprosy, 21 Lifestyle factors, 29, 74 Light, ultraviolet, 26 Lind, James, 38 Logistical factors in measurement, 113 Lumpectomy, 10–11 Lung cancer, smoking and, 130 M Malignant mesothelioma, 162 Mastectomy, total versus lumpectomy, 10–11 Match, frequency, 77 Matching, 57 Mean, regression toward, 80–81 Measles, historical study of, 23 Measurement, 111–155 of association, 111 with categoric variables, 126–131 with continuous variables, 145–149 bias in, 152–153 with categorical variables, 115–144 choice of variable for, 112 choosing form of, issues in, 112–114 with continuous variables, 145–153 dimensions of, 112–114 ethical issues in, 113 of frequency, 116–120 of impact, 120–126 importance of, 113 logistical factors in, 113 precision of, 113 sensitivity of, 113–114 of variables, 112–114 Medicaid coverage and use of ambulatory mental health care, 12 Mental health care, ambulatory, Medicaid coverage and use of, 12 Mesothelioma, malignant, 162 Methodology, ethics and, 182 Methods studies, Minimization, 55–57 6/30/09 10:52:12 PM Index Modified host factors, 27–29 Mortality, 120, 121 age-specific, 122–123 annual, 120, 121 and birth cohort, 27 proportional, 121–122 standardized, 123–127 calculations for, 124 Multiple sclerosis, 159–160, 165, 166 Myocardial infarction cut-point for test for, 139 oral contraceptive use and smoking in risk for, 78 N Nagasaki, National Diet-Heart Study, 69 Natural and Political Observations Made upon the Bills of Mortality, 22 Negative predictive value, 134–135, 137 Negative rate, false and true, 132–134 Neyman bias, 66–67, 152 Niacin deficiency, pellagra and, 26 NNT; see Number needed to treat Nocturnal enuresis and need for bifocals, relationship between, 75, 76 Nomenclature for epidemiologic research strategies, 86 Nominal variables, 114, 115 Nonrandom allocation, 58 Nonrandom sampling as form of bias, 65 Nuclear “accidents,” 7–8 Number needed to treat, 129 Nuremberg Code, 173 Nutritive agents, 26 O Obesity, 162 Observational studies, 40–41 Observations, number of, 41–42 Odds, relative, 130–131 Odds ratio, 131 On the Mode of Communication of Cholera, 24 Operating characteristic curves, receiver, 139–140 OR; see Odds ratio Oral contraceptive use and smoking in risk for heart attack, 78 Ordinal variables, 115 Osteoporosis, calcium deficiency and, 26 Outcome of research measuring, 41–42 nomenclature for, 86 Index.indd 195 195 P P(D), 138 P(T + |D), 138 P(T − |D), 138 P(T + |D), 138 P(T − |D), 138 Pandemic, 34 Panum, Peter Ludwig, 23 Passive surveillance, Patients, psychiatric, involuntary admission of, 41 Payment to physicians in research studies, 175–176 to research subjects, 176 Pearson product-moment correlation, 146–148 Pellagra, niacin deficiency and, 26 Period prevalence, 118–119 Person, 27–30 Physical agents, 26 Physicians in research studies, payment to, 175–176 Place, 30–31 Placebo controls, ethics of, 183–185 Placebo effect, 70 Plague(s) biblical, 21 bubonic, 25, 34 Plausibility, biologic, 165 Play the winner, 59 PMR; see Proportional mortality Poisons, 26 Policy analysis, 12 Population, 46 Positive predictive value, 134–135, 137 Positive rate, false and true, 133, 134 Predictive validity, 151 Predictive value positive and negative, 134–135, 137 prevalence and, relationship between, 135–136 Prevalence, 118 of HIV antibodies, 132, 136 incidence and, relation between, 119 period, 118–119 predictive value and, relationship between, 135–136 Probability sampling, 47–50 Product-moment correlation, Pearson, 146–148 Prolective studies, 43 Proportional mortality, 121–122 Prospective studies, 42–43 Protein deficiency, kwashiorkor and, 26 Protozoa, 27 6/30/09 10:52:12 PM 196 PDQ EVIDENCEBASED PRINCIPLES AND PRACTICE Proxy measures, 72–73 Psychiatric patients, involuntary admission of, 41 R Radiation ionizing, 26 risks of exposure to, 8–9 Random allocation, problems with, 58 Random error, 149, 153 Random sampling, 48–49 stratified, 49–50 strictly, 48 Randomization block, 54–55 other forms of, 58–63 regular, 60 Zelen, 59–61 Randomization device or scheme, 53 Randomized allocation, 52–54 Randomized controlled trial, 10, 15, 39, 61, 97, 180 Ratio, odds, 131 Ratio variables, 115 RCT; see Randomized controlled trial Reasoning, inductive, changes in, with age, 84 Receiver operating characteristic curves, 139–140 Reduction, absolute risk, 129 Regression toward mean, 80–81 Relative odds, 130–131 Relative risk, 127–128 Reliability, 146, 149–150 interobserver, 149 split-halves, 150 test-retest, 150 Research epidemiologic, stages of, 1–2 generalizability of results of, 53 outcome of, measuring, 41–42 Research ethics committees, 182 Research methodology, 37–105 design elements of, 40–46 history of, 37–40 other forms of randomization in, 58–63 sampling in, 46–52 subject allocation in, 52–58 threats to validity in, 63–85 Research strategies, epidemiologic, 85–98 Research studies descriptive or analytic designs for, 85, 87–91 epidemiologic, 86 nomenclature for, 86 experimental designs for, 85, 97–98 Index.indd 196 Research subjects coercion of, 174–175 payment to, 176 Respiratory cancers versus cancers from all other sites, 126 Retrolective studies, 43 Retrospective studies, 42–43 Retrospective-prospective studies, 43 Rickettsia, 27 Risk attributable, 129 with harmful exposure, assessing, 7–9 relative, 127–128 Risk reduction, absolute, 129 RO; see Relative odds ROC; see Receiver operating characteristic curves RR; see Relative risk S Sample, 47 Sampling, 46–52 adaptive, 59 cluster, 50–51 of convenience, 51–52 haphazard, 51–52 nonrandom, as form of bias, 65 probability, 47–50 random, 48–49 stratified, 49–50 strictly, 48 Sclerosis, multiple, 159–160, 165, 166 Scurvy, 38 Seasonal affective disorder, 32 Selection, subject, 52 biases in, 65–70 Sellafield, Sensitivity of measurement, 113–114, 134 Sickle cell anemia, 27 Single blind study, 70 Sleeping sickness, 30 Smoking lung cancer and, 130 and oral contraceptive use in risk for heart attack, 78 SMR; see Standardized mortality rate Snow, John, 23–24 Social desirability bias, 153 Special groups, informed consent by, 178–180 Specificity, 134, 135, 136 of association, 161–162 Split-halves reliability, 150 Standardized mortality rate, 123–125 calculations for, 124 Staphylococcus, Statistics, cancer, Canadian, 121 6/30/09 10:52:12 PM 197 Index Strata of key variables, 49 Stratified allocation, 55 Stratified random sampling, 49–50 Strength of association, 160 between variables, 126 Strictly random sampling, 48 Subject allocation nomenclature for, 86 in research methodology, 52–58 Subject selection, 52 biases in, 65–70 Suicide cyclical pattern of, 32 peak incidence of, 32 Surrogate end points, 72–73 Surrogate explanatory variables, 73 Surveillance, active and passive, 5–6 Survey, cross-sectional, 87 Suspicion bias, diagnostic, 74–75 Syndrome new, identifying cause of, 4–7 syndrome, 74 toxic shock, identifying, 4–7 Syndrome syndrome, 74 T Tay-Sachs disease, 27 Temporality of association, 162–163 Test(s) accuracy of, 141 diagnostic, 131–136, 142, 143 for myocardial infarction, cut-point for, 139 positive and negative, 133–134 sensitivity and specificity of, 134 Western blot, 133, 136 Test-retest reliability, 150 Testing, hypothesis, Thalassemia, 27 Th iamine deficiency, beriberi and, 26 Th reats to validity, 37, 63–85 Th ree Mile Island, Time, 31–32 Time course of research outcome, 41 Timing of research outcome, 41 Total mastectomy, 9–11 Toxic shock syndrome, identifying, 4–7 Transition, epidemiologic, Transitory host factors, 29 Treat, number needed to, 129 Treatment, determining effectiveness of, 9–11 Trial, controlled, randomized, 10, 15, 39, 61, 97, 180 True negative rate, 134 Index.indd 197 True positive rate, 133 Trypanosomiasis, tsetse fly and, 30 Tsetse fly, trypanosomiasis and, 30 TSS; see Toxic shock syndrome Tuberculosis, treatment-resistant, U Ultraviolet light, 26 Uncertainty, role of, 182–183 Undermatched groups, 57 Undermatching, 57 Urea formaldehyde foam insulation issue, 75 V Vaginal bleeding and endometrial cancer, association between, 67–68 Validity, 146, 150–152 concurrent, 151 construct, 151–152 content, 150 criterion, 151 face, 150 predictive, 151 threats to, 37, 63–85 Variable(s) categoric, 114–115 measurement with, 115–144 measures of association with, 126–131 choice of, for measurement, 112 continuous, 114–115 measurement with, 145–153 measures of association with, 145–149 dependent, 40, 72 explanatory, surrogate, 73 independent, 40, 72 interval, 115 measurement of, 112–114 nominal, 114, 115 ordinal, 115 ratio, 115 strength of association between, 126 types of, 114–115 Vectors, 34 Verbal consent, 180–181 Viruses, 26 Vitamin B1 deficiency, beriberi and, 26 Volunteer bias, 68–69 W Western blot test, 133, 136 Z Zelen randomization, 59–61 6/30/09 10:52:12 PM Index.indd 198 6/30/09 10:52:12 PM .. .PDQ* SERIES Also in this series: NORMAN, STREINER PDQ STATISTICS SCHLAGENHAUF-LAWLOR, FUNK-BAUMANN PDQ TRAVELER’S MALARIA MC KIBBON, WILCZYNSKI PDQ EVIDENCE-BASED PRINCIPLES... Contents Preface, xi Introduction to Epidemiology, What It Is, Trends in Epidemiology, Current Applications of Epidemiology, A Dose of Reality, 13 Classical Epidemiology, 21 A Little Bit of History,... xii 7/3/09 12:43:41 AM PDQ EPIDEMIOLOGY FM.indd xiii 7/3/09 12:43:41 AM FM.indd xiv 7/3/09 12:43:41 AM Introduction to Epidemiology WHAT IT IS Contrary to popular belief, epidemiology is not the