ABSTRACT
Besides investigations on etiology, epidemiology, and the evaluation of therapies, the identification and assessment of prognostic factors constitutes one of the major tasks in clinical cancer research. Studies on prognostic factors attempt to determine survival probabilities, or more generally, to predict the course of the disease for groups of patients defined by the values of prognostic factors, and to rank the relative importance of various factors. In contrast to therapeutic studies, however, where statistical principles and methods are well developed and generally accepted, this is not the case for the evaluation of prognostic factors. Although some efforts toward an
improvement of this situation have been undertaken,1-5 most of the studies investigating prognostic factors are based on historical data lacking precisely defined selection criteria. Furthermore, sample sizes are often far too small to serve as a basis for reliable results. As far as the statistical analysis is concerned, a proper multivariate analysis considering simultaneously the influence of various potential prognostic factors on overall or event-free survival of the patients is not always attempted. Missing values in some or all of the prognostic factors constitute a serious problem that is often underestimated.
