Statistical Analysis for validating and improving the staging system for breast cancer Open Access
Zhang, Yiran (Spring 2018)
This thesis project is aimed to utilize the National Cancer database (NCDB) to validate and improve the new breast cancer staging system proposed in the 8th edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual published in 2017. This staging system incorporates breast cancer biomarkers and will be widely used to determine the breast cancer prognosis worldwide. Our analyses were based on 420,520 breast cancer (BC) cases that were diagnosed from 2010 to 2014 and received the standard treatments. With the primary time-to-event outcome specified as time from diagnosis to all cause death, our univariate and multivariate survival analyses show that age, tumor grade, presence of lymph vascular invasion (LVI), hormonal receptor (HR) and HER2 status, and being triple negative breast cancer (TNBC) status, were significantly associated with the overall survival (all log rank test p-value<0.0001). We further identified that TNBC patients had worse overall survival times than non-TNBC , which included HR+/HER2+, HR+/HER2-, HR-/HER2+ in all stages and sub-stages (all p-value <0.0001). We constructed 4 different staging systems: stage + HR and HER2 status + age group + grade + LVI; stage + TNBC status + age group + grade + LVI; sub-stage + HR and HER2 status + age group + grade +LVI; sub-stage + TNBC status + age group + grade +LVI, and compared their performance based on the Harrell’s C-index, Uno’s C-statistics and Akaike’s information criterion (AIC). Our results indicated that the point system defined based on sub-stage + TNBC status + age + grade +LVI performed the best with the highest Harrell’s C-index (0.7316) and Uno’s C-statistics (0.6508) and the lowest AIC (488138.91). Our study also suggested that grouping breast cancer subjects by TNBC vs Non-TNBC has similar survival prognostic power to the more detailed BC classification based on HR/HER2 status. Our new staging system improves the prediction of all-cause survival over the traditional anatomic tumor, node and metastasis (TNM) system.
Table of Contents
II Patients and Methods
About this Master's Thesis
|Committee Chair / Thesis Advisor|
|Statistical Analysis for validating and improving the staging system for breast cancer ()||2018-04-09 14:00:57 -0400||