M-bioscore: proposing a new statistical model for prognostic factors in metastatic breast cancer patients

Abstract

Aim: The current study aims to propose and internally validate ‘M-bioscore’, which is a model to help predict the outcomes of untreated metastatic breast cancer patients. Methodology: Surveillance, epidemiology and end results (SEER) database (2010–2013) was accessed. Patients were divided into two groups: a training set and a validation set. Through a Cox proportional model, multivariate analysis for potential prognostic factors was performed. M-bioscore was calculated for all patients. Survival analyses were conducted through Kaplan–Meier analysis/log-rank testing. Results: A total of 6655 metastatic breast cancer patients were analyzed. In the training set, the following factors were linked to better cancer-specific survival in multivariate analysis: estrogen receptor positivity, isolated distant nodal metastases, progesterone receptor positivity, lower nuclear grade and HER2 neu positivity (p < 0.01). Cancer-specific survival was then assessed according to M-bioscore. Adjusted Cox regression cause-specific hazard (using breast cancer death as the event of interest) was evaluated in the validation cohort. Pairwise hazard ratio comparisons between different scores were significant (p < 0.05) except for the comparison between score 6 and 7. C-index for the validation cohort was 0.665 (Standard error (SE): 0.010; 95% CI: 0.646- 0.685).Conclusion: M-bioscore can predict the outcomes of untreated metastatic breast cancer patients. Validation of external datasets is needed.