A prediction model for metachronous colorectal cancer: development and validation

Area of Science:

• Oncology
• Epidemiology
• Biostatistics

Background:

• Estimating metachronous colorectal cancer (CRC) risk is crucial for personalized surveillance strategies.
• Current methods lack precision in identifying individuals at high risk for secondary CRC.
• Developing a robust risk prediction model can guide post-diagnosis management.

Purpose of the Study:

• To develop and validate a predictive model for estimating the 10-year risk of metachronous CRC.
• To identify key clinical and pathological factors associated with metachronous CRC development.

Main Methods:

• A large cohort of 6,085 population-based CRC cases diagnosed between 1997-2012 were analyzed.
• Cox regression with LASSO penalization identified predictors of metachronous CRC.
• Internal validation using bootstrapping assessed model discrimination and calibration.

Main Results:

• 138 cases (2.3%) developed metachronous CRC over a median of 12 years.
• Predictors included BMI, smoking, physical activity, family history, synchronous CRC, tumor stage, grade, histology, DNA mismatch repair status, and age at diagnosis.
• The model demonstrated good validity with a c-statistic of 0.65 and calibration slope of 0.873.

Conclusions:

• A validated risk prediction model can estimate individual 10-year metachronous CRC risk.
• The model incorporates readily available clinical variables for practical application.
• This tool can facilitate risk-stratified surveillance for colorectal cancer survivors.