Development of a novel malignant colorectal polyp prognostic nomogram

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Optimal treatment for malignant polyps involves balancing oncological risk against surgical morbidity and mortality.
• Uncertainty exists in selecting the best management strategy for these lesions.

Purpose of the Study:

• To develop an online risk-calculator utilizing machine learning.
• To predict the risk of adverse outcomes following polypectomy for malignant colorectal polyps.

Main Methods:

• A generalized linear mixed (GLM) model was developed using retrospective data from 2011-2020.
• Machine learning techniques were applied to a population-wide database of malignant polyps.
• Adverse outcomes included residual disease, lymphatic disease, or delayed recurrence.

Main Results:

• The final GLM model incorporated patient and pathological variables like age, gender, polyp location, invasion depth, and tumor grade.
• The model achieved an Area Under the Curve (AUC) of 0.76.
• The mean accuracy of the model was 0.76 (95% CI: 0.72-0.80).

Conclusions:

• A web-based nomogram was developed to aid clinical decision-making.
• The tool assists in determining whether patients require colorectal resection or can be managed with polypectomy.
• The risk calculator is available online at https://malignantpolyp.com/risk-calculator.