Survival Analysis: A Practical Approach

The seminal article by D.R. Cox proposing his namesake regression was published in 1972, making survival analysis a youngster in the family of statistics. It is a testimony to the field's success that the need for handbooks reviewing the current state of literature has developed. Survival Analysis is such a book, introducing the field in a manner that is both accessible and practical. The book moves in small, discrete steps, in a way, reminiscent of a survival curve itself! The authors spend two to three pages addressing a particular topic in non-technical style, followed by one or two examples from epidemiological trials. After a brief review of basic statistics, the books moves quickly into one-variable Kaplan-Meier survival curves. A discussion of the hazard rate ensues, followed by exponential and Weibell models in subsequent chapters. The authors then introduce two-variable and multi-variable models; it is approximately halfway through the book where the Cox proportional-hazards model is discussed. The remainder of the book concerns itself with practical aspects of the Cox regression. There is a chapter describing model-building strategies, including forward and backward stepwise models and the use of interaction-term variables. Special situations are also covered, such as time-dependent Cox models and the development of effective prognostic models. The book concludes with a catch-all chapter touching on applications of survival analysis as ranging from survival models of contrqception to meta-analysis. The break-neck pace with which the models and techniques are presented places the burden of statistical competency squarely upon on the reader's shoulders. Those who would benefit most from the book would have had at least one course in biostatistics at the graduate level, or a strong statistical background in non-epidemological fields. Survival Analysis is not for the reader with a casual interest in understanding survival curves; it is for practitioners of the art of model building. One peculiar quirk of the book is that the majority of its examples are from cancer trials in Great Britain, which may pose a difficulty for readers without access to the primary literature. While the non-technical nature of the material is a blessing to some, with its "soundbite" sections, Survival Analysis is not designed as a comprehensive textbook. The book is, however, the sort of book that one turns to before consulting other books. It is a place to refresh one's memory on a particular topic, or to use as a source of ideas, novel approaches, and technical solutions. As a supersaturated collection of techniques and examples accompanied by a full set of references, Survival Analysis will serve as an invaluable resource to those who engage in clinical trial design or statistical modeling .