Development of a Logistic Regression Model for Colorectal Cancer Relapse Prediction in Sabah, Malaysia

Area of Science:

• Oncology
• Biostatistics
• Public Health

Background:

• Colorectal cancer (CRC) is a leading cause of cancer death in Malaysia, with high relapse rates impacting survival.
• Current relapse surveillance for CRC lacks individualized schedules, potentially leading to suboptimal patient outcomes.

Purpose of the Study:

• To develop and validate an interpretable logistic regression model for predicting CRC relapse in Malaysian survivors.
• To utilize routinely available clinical and pathological variables for personalized risk stratification.

Main Methods:

• Retrospective case-control study using data from Malaysian hospital cancer registries (2015-2020).
• Multivariable logistic regression analysis of ten clinicopathological variables in patients with at least five years of follow-up.
• Model performance evaluated using 10-fold cross-validation, AUC, and Hosmer-Lemeshow tests.

Main Results:

• Six predictors significantly identified: tumor stage, lymphovascular invasion, perineural invasion, carcinoembryonic antigen level, tumor grade, and chemotherapy completeness.
• The final model showed strong discrimination (AUC = 0.85) and good calibration.
• Risk stratification into low, moderate, and high-risk categories was achieved.

Conclusions:

• A validated, clinically practical model for CRC relapse risk stratification in Malaysian survivors was developed.
• The model's reliance on routine data ensures scalability, especially in resource-limited settings.
• Integration with digital tools can support risk-adapted surveillance and improve long-term CRC outcomes in Malaysia.