https://orcid.org/0000-0003-4229-2462
-
Views
-
Cite
Cite
Donna Pauler Ankerst, Probability Modeling and Statistical Inference in Cancer Screening by Dongfeng Wu, Chapman and Hall/CRC Biostatistics Series, 2024, ISBN: 9781032513300 https://www.routledge.com/Probability-Modeling-and-Statistical-Inference-in-Cancer-Screening/Wu/p/book/9781032513300, Biometrics, 2025;, ujaf048, https://doi.org/10.1093/biomtc/ujaf048
Close - Share Icon Share
Extract
Screening for early detection is one of public health’s most potent weapons against cancer because cancers caught in the preclinical stage, before symptoms occur, pose a greater chance for cure. Toward this goal, large screening trials of asymptomatic populations are performed worldwide, such as the North American Prostate, Lung, Colorectal and Ovarian Cancer Screening Study or the WISDOM Personalized Breast Cancer Screening Study. Such expensive long-term large-scale studies test multiple screening regimes, such as annual mammogram versus risk-based screening for breast cancer. Information from screening trials is used as a basis for public health recommendations, such as the current US Preventive Services Task Force recommendation of biennial screening mammography for women aged 40-74 years. These recommendations emerge from rigorous analysis of the empirical evidence with consensus among interdisciplinary teams comprising clinicians, public health and economic scientists, and, last but not least, statisticians.
Determination of optimal cancer screening regimes, including onset ages, interval timings, and durations, is a statistical optimization problem to increase the number of cancers detected during the preclinical stage as opposed to the clinical stage. Empirical data for the optimization arise from the screening trials, but often require complementary registries, such as average population life expectancies. These are integrated into a likelihood based on probabilities of the observed events, which may be analyzed using either a frequentist or a Bayesian approach, and, as to be expected, can be quite complex. Until this book appeared, the methodologies have remained strewn across publications and individual book chapters. This book is the first to my knowledge to present these topics as an orderly textbook that includes detailed derivations. Applied exercises covering probability derivations, simulation, and data science topics, such as data table formatting, enhance the interactive learning aspect of the book. Solutions to some exercises are provided. As such, the book is a pragmatic research and education addition for its target audience comprising researchers with graduate-level mathematical, probability, and statistical expertise. At the same time, skipping over the formula derivations, the down-to-earth discussions and clearly presented results appeal to non-technical cancer researchers.
