In medical research there is a clear trend toward the use of larger, prospective studies of clinically significant outcomes, and there is an equal trend toward dealing with increasing numbers of variables. From such studies come the algorithms for making medical decisions - often referred to as evidence based medicine, and statistical analyses comprise a large part of this process. To keep pace with medical research and to even read its literature requires ever increasing understanding of statistical methods, and all too often published papers omit the basics. To assist in the understanding of these methods, we offer this course. It is intended for general as well as academic pathologists and for residents and fellows in pathology. Although minimal prior statistical knowledge and experience is necessary, we assume that the participants want to learn greater detail about common statistical methods and that they already have some experience reading statistical results in the medical literature.

The course will emphasize evaluation of tests used in either clinical pathology or anatomic pathology, and we will utilize examples or data from previously published papers. (The abstracts and citations of these papers will be mailed to registrants before the course.) Then, we will describe and illustrate how statistical methods interact with the data to yield useful answers to the practical questions. Altogether, the presentation will deal with the basic concepts of probability, the clinical probabilities of sensitivity, specificity, and the predictive values, Bayes' Rule, ROC curves, random variables, distribution functions, statistics and tests of hypothesis, type I and II errors, p values, the logistic regression model for binary outcomes and the Cox proportional hazard model for analysis of survival time. Much of the detail will be provided in the course syllabus, and both printed copy and a CD of the presented slides will be made available. We will plan to leave ample time for open discussion during the course.