Analysis of variance and covariance for colorectal adenocarcinomas in man as a logical prelude to "staging".

This study reports initial experience at the Ellis Fischel State Cancer Hospital (EFSCH), Columbia, Missouri, with the analysis of variance and covariance as applied to the ability to predict the lethality associated with colorectal neoplasms as influenced by tumor,host, and treatment variables considered individually and as multivariate clusters. The application of the methods to the clinical setting is based on the fact that each patient has one survival time. However, in the mathematical sense, each patient is composed of many different variables interacting simultaneously in highly complex ways to determine this uniqueness. The feasibility of determining the prognostic score or weight of many variables is illustrated. The combinatory effect of these scores can provide prognostic scores for clusters of variables providing a tool of much greater specificity than is provided by traditional staging systems such as the TNM.