Prognostic Groups by Tree-Based Partitioning and Data Refinement Methods

Authored by: Michael LeBlanc , John J. Crowley

Handbook of Statistics in Clinical Oncology

Print publication date:  March  2012
Online publication date:  March  2012

Print ISBN: 9781439862001
eBook ISBN: 9781439862018
Adobe ISBN:


 Download Chapter



The proportional hazards (PH) model of Cox (1972) has long been used to identify prognostic groups of patients by using the linear component of the model (prognostic index), or informally through counting up the number of poor prognostic factors corresponding to terms in the fitted model. However, the model does not directly lead to an easily interpretable description of patient prognostic groups. An alternative to using prognostic indices constructed from the PH model is a rule that can be expressed as simple logical combinations of covariate values. For example, an individual with some hypothetical type of cancer may have a poor prognosis if ((age ≥ 60) and (serum creatinine ≥ 2)) or (serum calcium < 5). This chapter presents two general classes of methodologies for constructing these logical rules for prognosis: (1) tree-based methods which partition the data into multiple prognostic groups and (2) peeling or extreme regression which both lead to sequential refinement the data into patient subsets with either very good or very poor prognosis.

Search for more...
Back to top

Use of cookies on this website

We are using cookies to provide statistics that help us give you the best experience of our site. You can find out more in our Privacy Policy. By continuing to use the site you are agreeing to our use of cookies.