Multivariate statistics in oncology: a review

In recent years multivariate analysis has been applied to the study of prognosis, diagnosis, and treatment in cancer patients. This form of analysis has provided investigators with a method to interpret increasingly complex data sets. In the current review, multivariate procedures such as the Cox proportional hazards model, logistic regression, and stepwise regression analysis are examined and compared with emphasis on the advantages and disadvantages of each procedure. Guidelines for the proper usage and interpretation of results are presented.