Date of Award

January 2021

Document Type

Open Access Thesis

Degree Name

Master of Public Health (MPH)


School of Public Health

First Advisor

Caroline Johnson


Colorectal cancer (CRC) is the third major cause of cancer-related deaths in the United States in 2020. Sex-related differences in CRC stage, prognosis, and metabolism have become increasingly popular in cancer research. Males have poorer survival for CRC, but females with right-sided colon cancer (RCC) have aberrant metabolism correlated with poor survival. Delay in knowing the condition of CRC in female patients would result in poor prognosis, which could be avoided by predicting prognostic outcomes. Random Survival Forest (RSF) is ideal for exploration and making predictions using metabolomics data with high dimension, strong collinearity, and heterogeneity, which CPH models could not efficiently address. In this retrospective study including 197 patients, we applied an RSF prediction method based on the backward selection algorithm in 5-year overall survival (OS) for 95 female CRC patients and validated its performance. We also investigated Cox proportional hazard models (CPH), lasso penalized Cox regression (Cox-Lasso), and Logistic Regression (LR) and compared their predictive performances. RSF using the backward selection algorithm showed the best performance with the C-index of the training and testing sets reaching 0.81(95% CI: 0.810-0.813) and 0.78 (95% CI: 0.776-0.777) respectively and identified the five most predictive metabolites for female 5-year OS: glutathione, citrulline, phosphoenolpyruvate, lysoPC (16:0), and asparagine. Accordingly, the backward selection algorithm-based Random Survival Forest model using tumor tissue metabolic profile is promising for predicting 5-year OS for female CRC patients. The results could be easily interpreted and applied in preventive medicine and precision medicine, guiding clinicians in choosing targeted treatments by sex for better survival and avoiding unnecessary treatments.


This is an Open Access Thesis.

Open Access

This Article is Open Access