Can Biomarkers Improve Ability of NPI in Risk Prediction? A Decision Tree Model Analysis

authors:

avatar Mohammad Reza Baneshi 1 , * , avatar P Warner 2 , avatar N Anderson 2 , avatar S Tovey 3 , avatar J Edwards 3 , avatar JMS Bartlett 4

Health School, Kerman Medical University, Department of Biostatistics and Epidemiology, Kerman, Iran; Centre for Population Health Sciences, University of Edinburgh, Teviot Place, Edinburgh, United Kingdom
Centre for Population Health Sciences, University of Edinburgh, Teviot Place, Edinburgh, United Kingdom
Section of Surgical and Translational Sciences, Division of Cancer Sciences and Molecular Pathology, Glasgow Royal Infirmary, Glasgow, United Kingdom
Endocrine Cancer Group, University of Edinburgh, Edinburgh Cancer Research Centre, Western General Hospital, Crewe Road South, Edinburgh, United Kingdom

how to cite: Baneshi M R , Warner P, Anderson N, Tovey S, Edwards J, et al. Can Biomarkers Improve Ability of NPI in Risk Prediction? A Decision Tree Model Analysis. Int J Cancer Manag. 2010;3(2):e80664.

Abstract

Background: The Nottingham Prognostic Index (NPI) is widely-used in the UK for risk stratification of breast cancer patients. This paper aims to evaluate the ability of this index to detect patients with sufficiently low risk of recurrence that they could be spared harsh treatments, and to construct an enhanced prognostic rule that integrates biomarkers with clinical variables to achieve better risk stratification.
 Methods: We undertook review of published studies of outcomes in risk groups derived by applying NPI, and report estimated event-free rates extracted from Then we analysed biological and clinical variables for 401 ER+ patients, to develop a Tree-based Survival Model (TSM), for risk prediction, and estimated event-free rates by resulting risk-groups, Kaplan-Meier (K-M) curves corresponding to TSM and NPI were plotted.
Results: We concluded that NPI does not distinguish low risk patients with a sufficiently high event-free rate to make it likely clinicians would decide treatments with potential harmful side effects can be avoided in that group. On the other hand, in the decision tree constructed, utilising 3 biomarkers, nodal status and tumour size, the 4 risk groups were clearly diverged in terms of event-free rates.
Conclusion: There is considerable potential for improved prognostic modelling by incorporation of biological variables into risk prediction. Whilst low risk patients identified by our TSM model could potentially avoid systemic treatment, higher risk patients might require additional treatment, including chemotherapy or other adjuvant treatment options. However, the decision tree model needs to be validated in a larger clinical trial cohort.

Fulltext

Full text is available in PDF.
 

© 2010, Author(s). This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/) which permits copy and redistribute the material just in noncommercial usages, provided the original work is properly cited.