Statistics for Dental Clinicians E-Book

Michael Glick

• Herausgeber: John Wiley & Sons
• Kategorie: Fachliteratur
• Sprache: Englisch

Seitenzahl: 473

Veröffentlichungsjahr: 2023

Beschreibung

STATISTICS FOR DENTAL CLINICIANS Enables clinicians to understand how biostatistics relate and apply to dental clinical practice Statistics for Dental Clinicians helps dental practitioners to understand and interpret the scientific literature and apply the concepts to their clinical practice. Written using clear, accessible language, the book breaks down complex statistical and study design principles and demonstrates how statistics can inform clinical practice. Chapters cover the basic building blocks of statistics, including clinical study designs, descriptive and inferential statistical concepts, and interpretation of study results, including differentiating between clinical and statistical significance. An extensive glossary of statistical terms, as well as graphs, figures, tables, and illustrations are included throughout to improve reader comprehension. Select readings accompany each chapter. Statistics for Dental Clinicians includes information on: * How to understand and interpret the scientific language used in the biomedical literature and statistical concepts that underlie evidence-based dentistry * What is statistics and why do we need it, and how to effectively apply study results to clinical practice * Understanding and interpreting standard deviations, standard errors, p-values, confidence intervals, sample sizes, correlations, survival analyses, probabilistic-based diagnosis, regression modeling, and patient-reported outcome measures * Understanding and interpreting absolute risks, relative risks and odds ratios, as well as randomized controlled trials, cohort studies, case-control studies, cross-sectional studies, meta-analysis, bias and confounding With comprehensive coverage of a broad topic, written using accessible language and shining light on statistical complexity often found in writings related to clinical topics, Statistics for Dental Clinicians is an essential guide for any dental practitioner wishing to improve their understanding of the biomedical literature.

Leseprobe

Statistics for Dental Clinicians

Michael Glick, DMD, FDS RCSEd

Professor and Executive Director

Department of Preventive and Restorative Sciences, and Department of Oral Medicine

Center for Integrative Global Oral Health

Cochrane Oral Health Collaborating Center

School of Dental Medicine

University of Pennsylvania

Philadelphia, PA, USA

Alonso Carrasco‐Labra, DDS, MSc, PhD

Associate Professor, Department of Preventive and Restorative Sciences

Center for Integrative Global Oral Health

Cochrane Oral Health Collaborating Center

School of Dental Medicine

University of Pennsylvania

Philadelphia, PA, USA

Olivia Urquhart, MPH

Instructor, Department of Preventive and Restorative Sciences

Center for Integrative Global Oral Health

Cochrane Oral Health Collaborating Center

School of Dental Medicine

University of Pennsylvania

Philadelphia, PA, USA

Copyright © 2024 by John Wiley & Sons, Inc. All rights reserved.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey.Published simultaneously in Canada.

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, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per‐copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750‐8400, fax (978) 750‐4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748‐6011, fax (201) 748‐6008, or online at http://www.wiley.com/go/permission.

Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762‐2974, outside the United States at (317) 572‐3993 or fax (317) 572‐4002.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging‐in‐Publication Data applied for[PB ISBN: 9781119810162]

Cover Design: WileyCover Image: © Michael Glick, Alonso Carrasco‐Labra, and Olivia Urquhart

To our past, present, and future students and mentors.

Preamble

During the past decades, we have had the opportunity to conduct research and publish in the biomedical literature. We have also had the opportunity to teach statistics and epidemiology to a range of academicians, clinicians, residents and students. Based on these experiences we have realized the need to better understand and interpret the scientific literature. This is more important today than ever before because of the ease of accessing the enormous number of scientific materials being published every day.

This book is not intended to teach readers how to conduct research or perform statistical analyses. Instead, this book is written for those who are interested in understanding and interpreting the biomedical literature to inform clinical practice. Twenty chapters cover basic concepts ranging from a general understanding of statistics to more complex topics, such as meta‐analysis and regression analysis.

We realize that one of the most challenging areas in statistics is its very specialized language and jargon. In this book every statistical concept is explained in plain English, both within the chapters and in a separate glossary. For ease, most of the italicized words in the chapters can also be found in the Glossary. We have also attempted to minimize formulas and equations within the chapters, but for those who want a more in‐depth understanding we have added an appendix where we provide equations and formulas for all the “numbers” we discuss in the chapters.

Other opportunities for the more curious reader are sections in most chapter entitled “a few additional notes,” which complement the text, as well as suggest additional topics for those who are interested in gaining a further understanding. Furthermore, selected readings accompany each chapter.

Every chapter also has “Pullout boxes.” These sections offer non‐scientific examples or analogies that we believe provide simple illustrations of concepts discussed in the chapters. We have also included numerous tables and illustrations to serve as aids to facilitate the interpretation and application of the discussed statistical concepts.

We hope this book will help clinicians to inform clinical practice for the benefit of our patients.

1What is statistics and why do we need it?

Paraphrasing from H. G. Wells’s Mankind in the Making (1903), where Wells wrote, “The time may not be very remote when it will be understood that for complete initiation as an efficient citizen of one of the new great complex worldwide States that are now developing, it is as necessary to be able to compute, to think in averages and maxima and minima, as it is now to be able to read and write,” S. S. Wilks, in his presidential address to the American Statistical Association, declared, “Statistical thinking will one day be as necessary for efficient citizenship as the ability to read and write.” Moving forward 70 years, this statement has never been more relevant or germane.

Claims are made constantly about miracle cures for cancer, diets that promise 20‐pound weight loss in one week, daily wine consumption either adding years to one’s life or ushering in an early death, medication A lowering blood pressure better than medication B, toothpaste X cleaning better than toothpaste Y, and so forth. What to believe? Being able to understand and interpret the basis for these and similar claims is an important skill that many of us lack but can easily learn.

We are inundated with a vast array of scientific publications, easily available to anyone with a computer and internet connection. However, more is not always better, as many articles present diverse and, at times, contradictory information that must be appraised and interpreted to inform clinical practice. Data by themselves are not very useful and must be analyzed to generate meaningful information, which then needs to be further contextualized to a particular clinical setting and a patient’s condition and need. An understanding of how data from scientific articles are generated, analyzed, presented, and interpreted has become essential to the informed clinician. This is where statistics comes in.

Statistics is essentially the scientific language the purpose of which is to describe, understand, and communicate scientific findings. An understanding of this language is needed to critically read, interpret, and understand the biomedical literature, with the ultimate goal of using scientific information for making informed decisions about etiology, diagnosis, prevention, treatment planning, prognosis, public policy, and much more. If statistics is a language, statistical concepts are the grammar (the rules of a language), while generation, organization, analysis, presentation of data, and the words used in this scientific language make up the syntax (the structure of sentences). Many statistical concepts can be similarly exemplified by using language as a comparison. For example, two important notions in statistics are confounders and bias. A confounder can be compared to a prejudice (known and unknown), while bias is similar to putting forth an argument but unintentionally using an incorrect definition (e.g., using usurping power when describing a democratic election).

This book does not cover how to apply and perform statistical analyses but instead focuses on the understanding and interpretation of statistical concepts used in clinical research that are meaningful to health care professionals. An understanding of statistics, like learning a new language, starts with a grasp of commonly encountered statistical terms and concepts.

One main obstacle to understanding statistics is the field’s specialized terminology. Words used in statistics do not always have the same meaning as in everyday language and must be interpreted correctly. For example, in statistics a population represents all members in a group of interest, while a sample is a representative subset of a population. The word risk, often used in everyday language to refer to something that reflects a potential threat, in statistics has neither a positive nor a negative connotation and represents simply the probability for an event, whether desirable or undesirable, to happen. Another example is the word “error,” which in statistical vernacular sometimes means “variation” but can also be used to illustrate a mistake. Statistical terminology is explained and defined throughout this book; most italicized words can be found in the glossary.

Statistics is a homonym, a word that is spelled and pronounced one way but has multiple meanings. Statistics is a discipline involved with the collection, organization, analysis, interpretation, and presentation of data (statistical model). Statistics, the discipline, has subdivisions, such as biostatistics, the application of statistical concepts and techniques to topics in the biomedical sciences. Statistics is also the method utilized in the discipline of statistics. Furthermore, statistics is the collection of data through statistical methods. Lastly, statistics can be numbers that are computed from data in a sample. Statistics from data in samples are represented by Latin letters and mathematical symbols; for example, (x‐bar) denotes a sample mean, and “s,” “Std Dev,” or “SD” denotes standard deviation. Data generated from or applied to a population are called parameters. Parameters are represented by Greek letters; for example, μ (mu) denotes the population mean and σ (sigma) signifies the population standard deviation.

Methods to summarize and analyze data in statistics are generally divided into two major categories—descriptive statistics and inferential statistics (Figure 1.1). Descriptive statistics (the category) is the analysis that helps describe and summarize statistics or parameters from samples and populations from which data have been collected. Examples of descriptive statistics (here statistics means data collected with statistical methods) include measures of central tendency (mean, median, or mode), measures of variability (e.g., standard deviation, range, variance), distributions of the data, and confidence intervals. As stated by Grimes and Schulz, “Descriptive studies often represent the first scientific toe in the water in new areas of inquiry. A fundamental element of descriptive reporting is a clear, specific, and measurable definition of the disease or condition in question. Like newspapers, good descriptive reporting answers the five basic W questions: who, what, why, when, where … and a sixth: so what?”1 By itself, descriptive statistics cannot be used to form a conclusion about associations or intervention effects (e.g., a difference in disease status as a result of a treatment) and can therefore not answer a research hypothesis. Descriptive statistics does not infer, i.e., cannot induce, what the population from which a sample was drawn may look like.

Figure 1.1 Descriptive and inferential statistics.

As it is rarely feasible to collect data on a particular aspect of interest from everyone in a population, such as mean body weight of all men in New York City, a representative sample of men is drawn from the population of interest; from the statistics generated by this sample, we can use special statistical methods to estimate what the “true” (actual) mean weight might be if we had measured all men in the population (all men in New York City). The method used to estimate a parameter, differences between parameters, or associations between parameters based on sample statistics is called inferential statistics. In inferential statistics sample data are used to induce (infer, derive) what an unknown population parameter, from which the represented sample has been drawn, may look like—we basically extrapolate from the sample data to make an estimated induction about the population. It is even possible to make a claim for statistics being the science that tells whether something we observe can be generalized or applied to a new or different but similar situation. Obviously, inferential statistics is what we are mostly interested in to inform clinical care, and statistics (the discipline) helps us quantify the uncertainty or certainty of our generalizations and the probability of making an incorrect or correct conclusion.

It is important to realize that statistics can never provide absolute certainty. For example, if we want absolute certainty of what the mean body weight is among all adult men in New York City, we need to weigh each individual and then calculate the mean weight. If we could do this, we would have no use for inferential statistics. However, if we draw a representative sample from all men in New York City, we must take into consideration many uncertainties. With this in mind, statistics has been described as the “science of learning from data, and of measuring, controlling, and communicating uncertainty; and it thereby provides the navigation essential for controlling the course of scientific and societal advances.”2 One major goal of statistics is to enable decision making in the face of uncertainty. How and why these decisions are made form the basis for being an informed clinician.

Uncertainties can be reduced by utilizing specific statistical methods. For example, to get a representative sample of the population, we select men for our samples in a randomized manner; that is, every member of a population will have an equal chance of being selected into our sample. Nonrepresentative samples would, for example, be a selection of men who have gathered to participate in a sumo wrestling competition or those who are easily accessible to the researchers because they live nearby (i.e., convenience sample). Obviously, these men do not form a representative sample of all men in New York City. Another way to reduce uncertainty would be to select more rather than fewer men for our representative sample—increase the sample size. Assume there are approximately 3 million men in New York City—if we chose 10,000 instead of 10 men for our sample, there would be a greater chance that the mean weight calculated for the larger sample would better represent the true mean weight.

Sampling is one of the most central and fundamental concepts underlying statistics. Accordingly, random sampling error or random sampling variation is an important concept to understand to correctly interpret scientific findings. Random sampling error is the name for differences or variations in characteristics of the sample and those of the general population from which the sample was drawn. Random sampling errors are always present and expected even in a random sampling process, as sample statistics can only approximate a population parameter. In other words, it is unlikely that sample characteristics, such as the mean body weight observed in a sample, will exactly reflect those of the population from where the sample was drawn, or that multiple samples from the same population will have the same mean weight. Another way of looking at sampling error is as an imperfect reflection of a parameter. Fortunately, there are statistical methods that can quantify this variation. The importance of variations in statistics is obvious when looking at the definition of statistics in the Medical Subject Headings (MeSH) thesaurus: “the science and art of collecting, summarizing, and analyzing data that are subject to random variation.”

Methods used to compare samples or make probability predictions about data require standardizing or normalizing the data. Imagine a course in statistics where a student must take three quizzes and one final exam. The first quiz is worth 10%, the second 20%, and the third 30%, while the final exam is worth 40% of the course grade. Thus, if we consider a maximum of 100 points for a perfect grade in the course, the first quiz would account for a maximum of 10 points, the second quiz 20 points, the third quiz 30 points, and the final exam 40 points. If a student scored 90, 85, 70, and 95 on the quizzes and the final exam, respectively, what would be their final grade? One way to figure this out is to normalize the grades, that is, to calculate the “weight, or the contribution” of each grade to the total grade. The first quiz could have provided a maximum of 10 points, the second quiz 20 points, the third quiz 30 points, and the final exam 40 points. In the case of the student, the first quiz contributed 9 points (a score of 90% out of a possible 10 points), the second quiz 17 points (a score of 85% out of a possible 20 points), the third quiz 21 points (70% of 30), and the final exam 38 points (95% of 40), for a total of 85 out of a possible 100 points. The student’s final grade would be 85. Although the student scored 90 on the first quiz and only 70 on the third quiz, the lower score on the third quiz contributed more to the final grade than the higher score on the first one. The described method normalized the exam scores by changing individual scores to a percentage of a total number of points. Essentially, the value of each score was converted from one unit (percentage of scores) to another unit (points). Normalizing, or standardizing, data is a key element utilized in statistical methods.

Beyond understanding what terms like random and errors represent in statistics, another often encountered term is distribution. A distribution is a representation of all possible values (or intervals) and the frequency (how often) of each value for a given variable. Such a distribution is often depicted graphically. Our example of mean body weight in a sample can be visualized by plotting each observed weight on the x‐axis and the frequency—how often a specific weight is observed—on the y‐axis (Figure 1.2). Each value can also be converted to a normalized score, such as a standard deviation (Figure 1.2). These normalized scores can also be represented as a distribution. There are different ways a distribution can be used. For example, one function of a distribution (a distribution function) can be used to calculate the probability of a variable’s observed value.

The best known distribution and the most frequently encountered in statistical applications in the biomedical sciences is the normal distribution. The graphic depiction of a normal distribution (Figure 1.2) is often referred to as a bell curve or Gaussian curve. A normal distribution includes 100% of all possible scores, where 50% of the scores are above the mean (to the right of the mean) and 50% are below (to the left of the mean). Amazingly, when scores are normalized (converted) into units of standard deviations, 1 standard deviation above and 1 standard deviation below the mean always include approximately 68% of all possible scores (34% above and 34% below the mean); 2 standard deviations above and below the mean include a tad more than 95% of all possible scores; and 3 standard deviations above and below the mean include more than 99% of all possible scores (Figure 1.2). A standard deviation is the average difference between each collected score and the mean that we would expect from random samples of equal size drawn from the same population (Chapter 3).

Figure 1.2 The Standard Normal Distribution.

Fundamental concepts throughout statistics are precision and accuracy (Figure 1.3). Precision is a function of replicability—how close to each other are replicated measurements—while accuracy is an estimation of how close the observed measurements are to the “truth” (the actual population parameter). Precision reflects random errors (variations among the values of a sample), while accuracy reflects systematic errors (errors that affect all values, due to, for example, inaccurate measurement tools, incorrect units, or other methodological flaws). Statistical methods can estimate both precision and accuracy, assisting in clinical decision making.

After data have been collected (gathering the data about the body weight of men in the sample), processed (entering the data into a format, e.g., a spreadsheet, to facilitate future tabulations), and explored (making appropriate measurements, such as frequency, mean, and variability), the data are analyzed for the final presentation of results. According to the Office of Research Integrity (ORI) at the U.S. Department of Health and Human Services, data analysis “is the process of systematically applying statistical and/or logical techniques to describe and illustrate, condense and recap, and evaluate data”3 and “provide[s] a way of drawing inductive inferences from data and distinguishing the signal (the phenomenon of interest) from the noise (statistical fluctuations) present in the data).”4 Which statistical analysis is chosen is based on what the investigator wants to explore and the type of study.

Figure 1.3 An analogy for precision and accuracy.Interpretation of a shooter’s precision and accuracy.

Precise and accurate

Precise but not accurate

Not precise but accurate

Not precise and not accurate

Statistical techniques are applied never in a vacuum but in the context of a study design. Biomedical studies can broadly be divided into two major categories—experimental studies, such as randomized controlled trials (Chapter 18), and observational studies, such as cross‐sectional studies, case‐control studies, and cohort studies (Chapters 15, 16, and 17). Designating a study as experimental or observational hinges on whether the investigator explicitly assigns an intervention or just observes a phenomenon without control of the administration of the intervention. For example, if an investigator wants to explore if a fluoride rinse (the intervention being studied) may reduce the incidence of caries, she may assign (i.e., select which study participants will and will not be assigned a fluoride mouth rinse) a fluoride mouth rinse to one group of study participants and a placebo rinse to another group of study participants, follow the participants for a specific period, and then compare the incidences of caries between the two groups. Another possibility is that the investigator could, instead of choosing to assign the intervention, enroll a group of individuals who are or have already been using a fluoride mouth rinse and compare the incidence of caries in this group to that in a group who are not or have not been using a fluoride mouth rinse. If the investigator assigned an intervention, the study is classified as an experimental study or a trial; if no intervention was assigned by the investigator, the study is classified as observational.

In general, most clinical studies assess frequencies of variables of interest, which form the basis for both descriptive and inferential statistics. Analyzing the body weight of men in New York City from a sample is one illustration of analyzing frequency data with the help of statistics. Another common and important concept is comparing frequency data between two or more samples or populations. Depending of what is being studied, comparisons can take place with an intervention (e.g., evaluate whether medication A is better than medication B, or measure if one dental implant causes more bone loss compared to another dental implant) or without an intervention. The former scenario occurs under the confines of an experimental or observational study and typically compares the frequency of an outcome (a factor being predicted) to an exposure (any factor or characteristic that may explain or predict an outcome). The comparison can, after conducting analysis, be presented as an absolute comparison (e.g., risk difference and mean difference) or as a relative comparison (e.g., risk ratio and odds ratio). These comparisons are usually referred to as measures of association (Chapter 2).

Before an observed study result is published, it behooves the investigator to consider the internal validity (or simply validity) and the external validity (or generalizability or applicability) of the study. The internal validity is an assessment of whether or how bias and confounders and random error may have influenced the study outcome beyond the exposure. External validity is a measure of the generalizability of the study results.

Studies often report statistically significant or statistically nonsignificant findings. Unfortunately, the concept of statistical significance is often misunderstood and inappropriately reported. A good understanding and interpretation of statistical significance as well as recognizing the difference between statistical significance and clinical significance are essential to inform clinical practice. Several chapters in this book address these topics (Chapters 5, 6, 20).

One of the most important aspects for the reader of the biomedical literature is avoiding being misled by reported research results. This is a difficult task and entails being able to critically appraise the literature.5 The following chapters arm the reader with important information on making informed clinical decisions, by providing a solid foundation for assessing the biomedical literature. However, it is beyond the scope of this book to highlight all possible misinterpretations and spins affecting the biomedical literature.6

Selected readings

Carrasco‐Labra A, Brignardello‐Petersen R, Glick M, Azarpazhooh A, Guyatt GH, eds.

How to Use Evidence‐Based Dental Practices to Improve Your Clinical Decision Making

. ADA Publishing; 2020.

Carrasco‐Labra A, Tampi M, Urquhart O, Howell S, Booth HA, Glick M. How to identify, interpret and apply the scientific literature to practice. In: Glick M, Greenberg MS, Lockhart PB, Challacombe S, eds.

Burket’s Oral Medicine

. 13th ed. Wiley‐Blackwell; 2021:1059–1079.

Cumming G.

Understanding the New Statistics: Effect Sizes, Confidence Intervals, and Meta‐Analysis

. Routledge; 2012.

Glick M, Greenberg BL. A march toward scientific literacy.

J Am Dent Assoc

.

2017; 148(8):543–545.

Greenberg BL, Glick M. Essential statistical and research design elements to help critically interpret the literature. In: Glick M, ed.

The Oral‐Systemic Health Connection. A Guide to Patient Care

. 2nd ed. Quintessence; 2019: 24–37.

Hahs‐Vaughn DL, Lomax RG.

An Introduction to Statistical Concepts

. 4th ed. Routledge; 2020.

Spiegelhalter DS.

The Art of Statistics: How to Learn from Data

. Basic Books; 2019.

Urdan TC.

Statistics in Plain English

. 5th ed. Routledge; 2022.

Notes

1

Grimes DA, Schulz KF . Descriptive studies: what they can and cannot do.

Lancet

.

2002; 359: 145–149.

2

Davidian M, Louis TA . Why statistics?

Science

.

2012; 6; 336(6077)): 12.

3

Office of Research Integrity, U.S. Department of Health and Human Services. Responsible conduct in data management.

Data Analysis

(

https://ori.hhs.gov/education/products/n_illinois_u/datamanagement/datopic.html

).

4

Shamoo AE, Resnik BR .

Responsible Conduct of Research

. Oxford University Press; 2003.

5

Carrasco‐Labra A, Brignardello‐Petersen R, Azarpazhooh A, Glick M, Guyatt GH . A practical approach to evidence‐based dentistry: X: How to avoid being misled by clinical studies' results in dentistry.

J Am Dent Assoc

.

2015; 146(12):919–924.

6

Boutron I, Ravaud P . Misinterpretation and distortion of research in biomedical literature.

PNAS

.

2018; 115(11): 2613–2619.