Abstract
The book is the fifth edition of this text. It is a good basic overview of epidemiology methods, though it refers to interpreting the medical literature. It is a useful guide for the health professional not trained in epidemiology. If you’ve worked with epidemiologists or read their reports, you know that there is much to read. Designs are complex and statistical analyses have become more sophisticated and nontransparent.
There are 14 chapters, which have some basic grouping. Chapters 2 to 6 cover issues of study design. Chapter 7 is about measuring or collecting observations. Chapters 8 and 9 are about data analysis. Chapters 10 to 12 cover Interpretation. Chapter 13 covers some outstanding issues not covered in the prior chapters, and Chapter 14 is a summary.
Throughout the book the author does a good job of using examples and illustrations to explain his message. This makes it easier to understand their importance.
The first chapter gives a somewhat elementary approach for keeping up with the literature. Techniques for reading articles and books are considered, which many of us learned in the 6th grade if not earlier. The author does highlight the trap that many in science fall into regarding keeping a pile of articles for which we have “good intentions” to read but which may not be relevant at the moment. Dump them.
For a toxicologist, Chapter 2 is quite useful if you’ve ever sat in a room with a group of epidemiologists and listened to them carry on about their work. Epidemiologists talk in code and, like most of us, have a unique way of disguising what they are doing using jargon in order, I think, to confuse toxicologists. The author breaks down the elements of epidemiology studies and gives clues regarding how to determine what they did regardless of what they call it.
In Chapter 3 the author gives an excellent overview of the various approaches to case control studies. There are a number of problems with case-control designs and the author goes through the pitfalls and provides guidance regarding how to tease these out. There is a good discussion regarding the selection of controls and important factors to consider. Also, exposure information is critical and needs to be carefully assessed. Finally, he discusses the issue of bias and how it can enter the study design and how it should be controlled.
A good overview and use of simple examples to explain the nuances of cross-sectional and follow-up studies is given in Chapter 4. These studies tackle the task of providing explanations by assembling groups of subjects that represent either a general, nondiseased population of people who share features we think might predispose them to a particular outcome. He points out pitfalls and the importance of sampling selection. The question that readers must ask is whether the sampling technique employed has guarded against the selection of a biased or unrepresentative sample. Authors may report that patients were selected at random when, in fact, strict randomizing techniques were not used.
The author explains that ecologic studies are cross-sectional in structure and many would argue that these studies are only useful for hypothesis generating. He asserts though that they are highly efficient and economical.
The follow-up, or cohort, design is generally considered to be the most desired of observational methodologies. Loss to follow-up studies is a critical drawback. The author points out several ways that investigators can deal with the follow-up problem and gives advice on how to check behind the author to see if this homework has been performed. The chapter concludes with a number of helpful questions to identify if a study design is a cross-sectional or follow-up approach.
In Chapter 5, the “controlled trial” is discussed. This design may be more familiar to those in pharmaceutical research. These studies are often reported in the press and understanding the issues of these studies can be helpful for general conversation as well as potential litigation. Controlled trials have associated tribulations that we should be aware of.
The author covers four major areas of methodological concern: enrollment, allocation, follow-up, and analysis. A good use of tables and illustrations for examples are helpful in understanding the issues of these studies. Many of the issues discussed for other studies apply here as well such as “randomization, blinding and attrition.”
The randomized control trial is not always the study design of choice and the reasons are discussed in Chapter 6. Some of the reasons include real-world considerations, practical constraints, and ethical concerns. The author goes through each of these and uses good examples to make the points. One assertion is that “sometimes the act of conducting a trail runs into ethical concerns.” I suggest that this is always the case and “ethical concerns” should be addressed in every study design involving human subjects, regardless.
Chapter 7 is dedicated to the collection of data and pitfalls to be aware of. Even to the toxicologist, data collection is not without error. The author reviews this imperfect process and gives pointers for what to watch for in human studies. Sources of error can include variability in recording data as well as investigator and subject bias. Keys are given for reducing measurement error such as blinding, establishing unambiguous standards, validating instruments, and supervised training for data collectors.
In Chapter 8, the author gives a basic overview of some of the issues related to the application of statistics to aid in data interpretation. He discusses some of the pitfalls and offers good advice about reliance on “statistical significance” alone for data interpretation. Principles underlying the use of tests of statistical significance are explored, as well as terminology and review of some common, subtle “statistical traps.” Results may be statistically significant but practically unimportant as well as differences that are not statistically significant are not necessarily unimportant.
In Chapters 9 and 10, additional time is given to statistical tests and interpretation. These chapters are a must for all toxicologists. The author points out that understanding the principles that guide the selection and application of statistical tests aids critical reading. One significant point is that the principles of sound study design still apply and no statistical methodology will overcome poor design.
Many of the principles laid out are so often overlooked, ignored, forgotten or considered a hassle. In toxicology, many new test methods have been introduced (e.g., developmental neurotoxicity, immunotoxicity methods) and many will be introduced as alternatives, which would benefit from the questions put forth in Chapter 10. They relate to the concepts of sensitivity, specificity, and predictive value. These can be difficult issues to answer in toxicology due to lack of information and the issue of species extrapolation. However, the effort should be made because it would greatly aid in decision making when trying to decide the value of doing a test and identifying areas of uncertainty. We should also try to factor in the frequency, or prevalence, of a condition or outcome in the population as part of the decision. This has been the subject of debate in the field of alternatives to animal testing and in the prescriptive approach taken in REACH guidance documents for information requirements.
The author concludes by recommending that when articles offer information on new diagnostic tests, new application of old tests, or novel ways of utilizing clinical symptoms and signs to identify diseases, ask the following questions: Have sensitivity, specificity, and predictive values been calculated? Do the authors give evidence that they understand the importance of misclassifying patients into disease (false-positive) or nondisease (false-negative) categories? Has the predictive value been correctly ascertained with reference to the prevalence of the disease in the population? Finally, does the test improve on the present state of affairs? Is it more accurate, less costly, less painful, less time consuming, or in some other way better than the diagnostic techniques currently in practice? All are excellent questions to act as responsible investigators.
In Chapter 11, a good overview is presented for the various ways to express risk and this is important to understand. Relative risks are often reported in the literature and the reader needs to understand this in relation to the type of study being reported. The difference between “relative risk” and “attributable risk” can have implications for decision making. Balancing risks and cost/benefits can be difficult topics but are important for policy makers as well as individuals in making choices. Risk statements must be kept in perspective.
Chapter 12 deals with the issue of attributing cause based on the findings from studies. The classic publication of Hill is presented where he describes nine features of relationships that are useful in constructing a case for causation. Particularly issues of strength of association, dose-response, biological plausibility, and consistency are discussed.
The book concludes with a general discussion of other types of reports such as editorials, letters, and reviews. This is a useful reference and helpful guide for health professionals trying to understand the medical literature and/or epidemiology.