 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
| |
|
||||||||||||||||
| |
|
||||||||||||||||
|
|
Book Review Quantitative Methods for the Evaluation of Cancer Screening
Edited by Stephen W Duffy, Catherine Hill and Jacques Estève New York (US): Oxford University Press, 2001; Evaluation of cancer screening is a complex task with many pitfalls, even for the informed practitioner. This unique and much-needed volume not only systematically compiles diverse advanced quantitative methods normally scattered across the scientific literature, but also considers both their theory and practice in applied settings. It includes methods for evaluating cancer screening at various steps of program planning, development, implementation and monitoring. These approaches may be used to determine, first, whether screening works (or efficacy); second, whether to establish a program, often in a manner that reflects understanding of tumour biology (effectiveness); third, how to modify it commensurate with both biology and available resources (cost-effectiveness); and fourth, whether the program is having the desired impact (surveillance and evaluation). Editors Steven Duffy, Catherine Hill and Jacques Estève have skillfully collated the contributions of the various authors based on material from a workshop held in Paris in 1997. The fourteen chapters, generally of very high quality, include study designs for measuring the efficacy and the effectiveness of cancer screening, advanced quantitative methods such as Markov chain models, use of routine data, and cost effectiveness. While most examples focus on breast and cervical cancer, some material is presented on colorectal cancer and childhood neuroblastoma. Chapter 1 concisely summarizes approaches useful in evaluating cancer screening. These range from the rigorous examination of the criteria of Wilson and Jungner needed prior to implementing a screening program; through the role of the randomized controlled trial (RCT); then to the need for early indicators of program effectiveness, as mortality reductions may occur too far in the future to be useful. Concepts essential for an understanding of cancer screening techniques, including study results and lead time, are well illustrated with clear tables and figures. The summary of developments in the field over the past four decades provides a framework for study design in evaluation by emphasizing the importance of the RCT in establishing the efficacy of screening, then describing designs to evaluate programs already in place, including the uncontrolled cohort design and the retrospective case control study. The excellent overview of the important influences of effectiveness and costs on cancer screening evaluation given in Chapter 2 highlights the large amount of data required, including epidemiology in the absence of screening, demography, screening quality and policies, clinical practice and the costs involved. Chapter 3 covers the issues of contamination and compliance in screening trials; the discussion on how to improve compliance in those invited for screening is relevant to both RCT design and to ongoing program operations. Statistical methods to adjust for contamination and non-compliance allow for an unbiased assessment of screening effectiveness yet still respect randomization, and will be useful to those assessing RCT results in the literature. Chapter 4 presents a more specialized analytic design to reduce selection bias in the absence of randomized controls by estimating the reduction in cancer mortality caused by screening at a younger age. The next three chapters consider statistical models that evaluate various aspects of screening, based on improved understanding of the biology of the disease (in all cases with breast cancer as an example). The use of Markov chain models to estimate rates of disease progression, including estimates of sojourn time, or pre-clinical detection period, is described in depth in Chapter 5, and includes several pages of valuable SAS computer code. A statistical method that describes the theoretical relationship at the population level between the rate of clinically detectable metastases at diagnosis to the rates of occult metastases is presented in Chapter 6 with supporting data, figures and statistical appendices. Another Markov chain model, described briefly in Chapter 7, predicts the mortality reductions likely to accrue from different screening intervals by estimating the rates of development of preclinical breast cancer, progression from preclinical to clinical disease, and progression from clinical disease to death. Practical considerations and the use of routine data for evaluation are discussed in Chapters 8 to 10. An excellent discussion of the use of the source data needed in evaluation, including cancer registry information, is accompanied by the many ways cancers can be categorized by detection mode. This is must reading for those analyzing interval cancers. The pitfalls of cohort and case-control design are discussed next, particularly as applied to screening exposure histories for cervical cancer. Finally, Hakolinen elegantly describes how to monitor the impact of screening using incidence and mortality data by utilizing methods that range from simple trends analysis or comparison of age-specific rates in different periods, to cancer projections, to multilevel analysis. Readers will gain many useful insights into confounding factors such as treatment advances on mortality and changes in diagnostic practice and risk factor prevalence on incidence. The final four chapters look at the impact of screening on incidence and mortality for cervical, breast, and colorectal cancers, and childhood neuroblastoma. Walter analyzes enhanced data from the original Canadian case-control study on cervical cancer to calculate pre-clinical detection periods (PCDP), finding that the most important benefits accrue by detecting disease before it becomes invasive. The high sensitivity and long PCDP means that screening as seldom as once in 10 years is effective, and underscores the importance of recruiting women for the initial screen. Paci and colleagues next develop simple methods for estimating the extent of possible over-diagnosis of breast cancer and predicting mortality benefits from screening that will be useful as the program evolves to ensure that observed incidence and mortality counts are on track. A method to estimate the sensitivity of the fecal occult blood test (FOBT) for colorectal cancer that also accounts for the mean sojourn time allows for varying these estimates across the proximal and distal colon and the rectum. In the final chapter, Estève et al. develop models to assess screening strategies for neuroblastoma. Their analysis illustrates the importance of synthesizing biological, genetic, and epidemiologic knowledge into models; such wisdom would have been useful prior to implementing large-scale screening programs and efficacy trials and has contributed to the current consensus against screening or further trials. The workshop nature of the original material used in the book limits its scope, as not all relevant topics are covered. The depth and breadth of coverage varies: while in most chapters the reader is guided to relevant literature that will fill any gaps, for some topics essential literature, such as the selection of control groups for case-control studies, is missing. A glossary of terms would help readers new to the field to understand that pre-clinical detection period and sojourn time are one and the same, for example. More importantly, while the book includes an excellent discussion of Markov type models together with many creative suggestions for their adaptation, there is almost no discussion of other modeling approaches used in the literature, such as simplified life-tables that can be based on spreadsheets, decision analysis approaches, or more comprehensive population-based micro and macro-simulation models. Finally, given the title of the book, readers may expect more information than is included on interim evaluation indicators essential to the ongoing monitoring of screening programs to ensure high quality service and cost-effectiveness. Despite these caveats, this book will be an essential reference to serious evaluators of screening programs and policies. Applications are generally described in enough technical detail for a reader to try the approaches and supported by details of calculations, and, in some cases, computer code. This volume meets its aim to be accessible to a range of professionals in public health and disease control, as well to statisticians and epidemiologists.
Leslie Gaudette Acting Chief, Treatment and Outcomes Surveillance
[Previous] [Table
of Contents] [Next]
|
 |
|||
|
Last Updated: 2002-02-21 |  |
|
