A prognostic model for advanced colorectal neoplasia recurrence

Area of Science:

• Gastroenterology
• Medical Statistics
• Preventive Medicine

Background:

• Current US Multisociety Task Force (USMSTF) guidelines for post-polypectomy risk stratification have moderate accuracy.
• Identifying patients at high risk for subsequent advanced neoplasia (AN) is crucial for appropriate surveillance intervals.

Purpose of the Study:

• To develop and validate a statistical prediction model to improve the identification of patients at high risk for metachronous advanced neoplasia (AN) after colonoscopic polypectomy.
• To compare the performance of the new model against existing USMSTF guidelines.

Main Methods:

• Pooled data from seven prospective studies (n=8,228) with 3-5 year follow-up for metachronous AN.
• Data randomly split into training (n=5,483) and validation (n=2,745) sets.
• Developed a prognostic model using polyp location, prior history, patient age, and polyp characteristics; identified two risk cut-points.

Main Results:

• The final model incorporated polyp location, prior polyp history, patient age, and number, size, and histology of resected polyps.
• One risk cut-point improved sensitivity, while a second improved specificity (46.2% vs 42.1%) compared to USMSTF guidelines.
• The model achieved an estimated AUC of 65% (95% CI: 62-69%).

Conclusions:

• A model-based approach offers flexibility in balancing sensitivity and specificity for colonoscopy surveillance.
• Current clinical data provide only modest improvements in prognostic power.
• Further research incorporating factors like adenoma detection rate is warranted for enhanced risk prediction.