Analysis of incidence rates

This book is an excellent resource for graduate students, researchers, and analysts in epidemiology, biostatistics, social sciences, economics, and psychology. The author of this book, Dr. Cummings, is Professor Emeritus, Department of Epidemiology, School of Public Health, University of Washington, Seattle, WA. His research was primarily in the field of injuries. The focus of the book is to offer the most useful analysis methods for incidence rates, which is the event count occurred during the given person-time follow-up. Important pitfalls and areas of controversy are also discussed. Before becoming a professor for epidemiology, Dr. Cummings was a physician for 20 years. He used many practical examples in the book to illustrate the basic concepts of incidence rates and analysis methods, the differences, and interpretations. Many epidemiological or social example data from real practice in history are used to illustrate the concept and different views.

In the first half of the book, the author discussed the fundamental concepts of rates with an emphasis on topics not covered by other textbooks. Specifically, Chapter 1 is an introduction; Chapters 2–5 clarify the basic concepts such as incidence rates, standardized rates, risks, hazard ratios, risk ratios, odds ratios, and rate ratios. A Poisson process is assumed for the counting in intervals of time. Chapters 6–8 show stratified methods, including standardization, inverse-variance weighting, and Mantel–Haenszel methods. Problems related to collapsibility are described. From Chapters 9–13, Poisson regression and linear regression model fitting are thoroughly discussed. In the second half of the book, extensions or modifications of the Poisson regression model are covered; this includes the negative binomial regression, methods for clustered data, longitudinal analysis, Bayesian method, and the exact method. In Chapter 20, the matched cohort methods are presented in detail. Dr. Cummings used this method to estimate how the use of seat belts and the presence of airbags were related to death in a traffic crash. The last chapter of the book compares Poisson regression with the Cox hazard model. Also, Royston–Parmar models are introduced.

Apparently, Dr. Cummings is very experienced in using the Stata statistical software, and all example data presented in this book are analyzed using Stata. Stata commands are provided, and file details are available online. Stata is a commonly used application tool for researchers in epidemiology, economics, and social science. In the pharmaceutical field, SAS or R may be widely used, but they should all generate the same analysis results.

Incidence rates are commonly analyzed and compared in the pharmaceutical industry, especially in the clinical trials. For example, a test treatment is compared with placebo in terms of a safety incidence of interest such as headache and hypertension. When the patients on test treatment tend to have longer follow-up than placebo, an exposure-adjusted incidence rate is more meaningful. Liu et al. (2006) discussed several methods of calculating confidence interval for the difference of two incidence rates, in which Miettinen and Nurminen (MN) method was found to outperform the other methods, especially when the numbers of events are small.

To accurately assess or estimate the incidence rates or risk ratios, a large database with long follow-up is preferred. Such a large database may come from different sources. In recent years, numerous meta-analyses are published by pooling the data from different world regions and countries to study the trend of incidence for a specific disease or event of interest. A pooled risk ratio is derived and analyzed by sex and age groups (Li et al. 2018). Sex- and site-specific all-ages new cancer incidences and cancer deaths are estimated, and uncertainty intervals to take into account possible sources of bias are provided (Ferlay et al. 2018). These publications reveal the recent analyses and practice based on incidence or incidence rates.

In summary, Analysis of Incidence Rates is an excellent textbook, with extensive applied example data, provides fundamental concepts what incidence rates are, reviews their advantages and limitations, promotes understandings of analytic methods, provides practical suggestions for analyses, and points out problems and pitfalls.